University of Southern California Dissertations and Theses

Essays on the new economic history of the Middle East

(USC Thesis Other)

Essays on the new economic history of the Middle East

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

ESSAYS ON THE NEW ECONOMIC HISTORY OF THE MIDDLE EAST

by

Mohamed Saleh

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)

May 2012

Copyright 2012 Mohamed Saleh
ii

Acknowledgements

I am deeply indebted and grateful to my advisor, Dora Costa, who gave me her time,
support, encouragement, and advice since I started this dissertation more than three years
ago. Dora agreed to advise me on the dissertation despite the fact that I am a student at a
different institution. Since then, she has been available to meet me on a weekly basis,
provided me with her close guidance, integrated me into the economic history group at
UCLA, and helped me tremendously both before and on the job market. I believe that if it
were not for her amazing support and guidance, I would not have been able to pursue
economic history, which is my true research passion, as my main research field, nor
would I have been able to succeed in this field.
I am also deeply indebted and grateful to Leah Boustan, who helped me a lot with
her valuable comments and advice whether in our bi-weekly joint meetings or in the
quarterly UCLA Economic History Proseminars. I benefited a lot from her advice along
the way of developing and shaping the dissertation.
I am equally indebted and grateful to Jeffrey Nugent, whose support, kindness, and
guidance were amazing since I started the MA program at USC back in 2005 and
throughout the PhD program (2006-12). Jeff has always been such a great mentor,
teacher, and friend for all the students in the department.
I owe special thanks to Naomi Lamoreaux, for her valuable comments and her help
on the job market; Steven Ruggles, for his technical advice on the data collection, for
approving to support my project through IPUMS, and for his help on the job market;
iii

Ragui Assaad, for his very helpful advice on the data collection and the dissertation as
well for his help in introducing me to IPUMS; Richard Easterlin for his valuable
comments on the dissertation, his approval to be on my dissertation committee, and for
providing me with a Research Assistantship in my job market year. I am very grateful to
Hany Takla, who provided me with a wealth of historical resources through St. Shenouda
the Archimandrite Coptic Society in Los Angeles. I also want to thank Roger Owen,
Kenneth Cuno, and Khaled Fahmy, who all provided me with experts’ advice on Middle
Eastern history. Special thanks go to Donald Miller for agreeing to be on my dissertation
committee.
My data collection efforts would not have come to fruition in the absence of funding.
In this regard, the grants that I received from IPUMS, Economic History Association,
USC College of Letters, Arts, and Sciences, and Center for Religion and Civic Culture at
USC were particularly generous and essential for the success of the project.
The administration of the National Archives of Egypt was very cooperative and
helpful to allow me to undertake the data collection project over there, as well as to
provide me with equipments and data entry assistants. I am very grateful in this regard to
Saber Arab, Rifaat Hilal, and Emad Hilal for their great help and support. I was also very
lucky to work with a great team of data entry and research assistants. Among these I
would like to thank Hesham Shibl, Manal Zahran, Iman Sami, Caroline Naguib, and
Marina Anis for their great research assistance. I owe special thanks to Julie Iskander
whose great IT help was crucial for the success of the project, and to Audrey Dridi who
iv

was so kind to share with me many of the historical resources that she has gathered for
her MA thesis.
I would like to sincerely thank Dimitris Pipinis who has been such a great friend to
me since I came to USC, and whose insightful comments, discussions, and support were
so helpful along the way.
Last but not least, I would like to deeply thank my mother, who will forever be in my
mind, my father, whose great library and broad culture made me what I am, my siblings,
and my partner, Rania, for being in my life.

v

Table of Contents

Acknowledgements ii

List of Tables viii

List of Figures xi

Abstract xiii

Introductory Chapter 1

Chapter 1: A Pre-Modern Middle-Eastern Population Brought to Light:
Digitization of the 1848 and 1868 Egyptian Individual-Level Census
Records
11
1.1. Introduction 11
1.2. The Census Registers: Historical Context, Format of the
Registers, and Enumeration Methodology
15
1.2.1. Historical Context 15
1.2.2. Format of the Census Registers 18
1.2.3. Enumeration Methodology 23
1.3. Digitization Project and Sampling Strategy 24
1.3.1. Target Population and Sampling Frame 25
1.3.2. Sampling Mechanism 29
1.3.3. Comparison to the IPUMS 1850 U.S. Census Sample
and Alleaume and Fargues 1848 Egyptian Census Sample
36
1.4. Application: Exploring the Nineteenth Century Egyptian
Modernization Program
38
1.4.1. Descriptive Statistics 39
1.4.2. Application 40
1.5. Conclusion 45

Chapter 2: The Reluctant Transformation: Modernization, Religion, and
Human Capital in Nineteenth Century Egypt
55
2.1. Introduction 55
2.2. Historical Background 60
2.2.1. Guilds and State Production Projects 60
2.2.2. Kuttabs and Modern Schools 63
2.2.3. National Regular Army and Military Reforms 65
2.3. Data 66
2.4. Empirical Analysis: State Industrialization 69
vi

2.4.1. Technology-Skill Complementarity of State Production
Projects
71
2.4.2. State Industrialization, Religion, and Occupational
Differentials
73
2.4.3. Mechanisms of State Industrialization: Technological
Change or Job Training?
75
2.4.4. Examining the Validity of the Empirical Strategy 78
2.4.5. State Industrialization and Religious Occupational
Segregation
80
2.5. Empirical Analysis: Educational and Military Modernization 82
2.5.1. Educational Modernization 82
2.5.2. Military Modernization 85
2.6. Conclusion 87

Chapter 3: From Kuttabs to Schools: Educational Modernization,
Religion, and Human Capital in Twentieth Century Egypt
102
3.1. Introduction 102
3.2. History of Education in Twentieth Century Egypt: An Overview 107
3.3. Data 111
3.4. Empirical Strategy and Results 115
3.4.1. Evolution of Inter-Religious Differentials in Educational
and Occupational Outcomes Over the Twentieth Century
116
3.4.2. Impact of Expansion of Modern Schooling on Inter-
Religious Literacy Gap
119
3.5. Conclusion 122

Chapter 4: On the Road to Heaven: Poll Tax, Religion, and Human
Capital in Medieval and Modern Egypt
135
4.1. Introduction 135
4.2. Historical Background: Christianization and Islamization of
Egypt
140
4.3. Data 143
4.4. Empirical Analysis: Coptic Population Share and the Coptic-
Muslim Gap
146
4.5. Empirical Analysis: Historical Origins of the Coptic Spatial
Distribution
148
4.5.1. Islamic Poll Tax Institution: Historical Background 148
4.5.2. Eighth and Ninth Centuries Tax Revolts 152
4.5.3. Seventh – Ninth Centuries Arab Immigration Waves 153
4.5.4. Coptic Churches and Monasteries in the Twelfth and
Fifteenth Centuries
156
4.5.5. The Flight of the Holy Family to Egypt 158
4.5.6. Which Historical Origins Are More Important? 159
4.6. Understanding the Mechanisms 159
vii

4.6.1. When Did the Coptic-Muslim Gap Emerge? 159
4.6.2. Why Did the Gap Persist? 164
4.6.3. Other Historical Explanations for the Coptic-Muslim Gap 164
4.7. Conclusion 166

Bibliography 180

Appendices 189
Appendix A: Scanned Pages of the Census Registers 189
A.1. A Page from an Urban Province (Cairo) 189
A.2. A Page from a Rural Province (Al-Gharbiya) 190
Appendix B: Actual Examples of the Digitized Census Records 191
Appendix C: Data Appendix to Chapter 2 192
C.1. Coding the Occupational Titles in the Egyptian 1848 and
1868 Census Samples and Constructing the Social Status Index
(SSI)
192
C.2. Constructing the Modernization Index 197
C.3. Constructing the Industrial Affiliation Variable 199
C.4. Other Variables in the Empirical Analysis 199
Appendix D: Data Appendix to Chapter 4 204

viii

List of Tables

Table 1.1: Evaluating Coverage in the Rural Provinces in 1848 and 1868 47

Table 1.2: Estimating Individual-Level Coverage Rate in Each Province
in 1848 and 1868
48

Table 1.3: Duplicates: Geographic Units that Were Enumerated More
than Once by Province
49

Table 1.4: Sampling in 1848 and 1868: Geographic Distribution by
Province and Targeted and Actual Sampling Rates
50

Table 1.5: Probit Model for Probability of Coverage in Rural Provinces
in the 1848 and 1868 Censuses
51

Table 1.6: Selected Descriptive Statistics from the 1848 and 1868
Samples
52

Table 1.7: Sectoral Distribution of Adult Active Male Workers in 1848
and 1868 Egypt
53

Table 1.8: Occupational Distribution of Workers in the Modern Sector in
1848 and 1868
54

Table 2.1: Descriptive Statistics: Modernization, Religion, and Human
Capital in 1848 and 1868
93

Table 2.2: Testing for Technology-Skill Complementarity in State
Production Projects
94

Table 2.3: Impact of State Industrialization on Inter-religious
Occupational Differentials- Ordered Logit Model
95

Table 2.4: Impact of State Industrialization on Inter-religious
Occupational Differentials- Multinomial Logit Model
96

Table 2.5: Mechanisms of State Industrialization: Technology or Job
Training?
97

Table 2.6: Crowding-out Effects of State Industrialization 98
ix

Table 2.7: Occupational Segregation in Traditional and Modern
Production Sectors
98

Table 2.8: Educational and Military Reforms and Religious
Occupational Differentials
99

Table 2.9: Religious Composition of Religious and Modern Schools in
1848 and 1868
100

Table 2.10: The Composition of the Egyptian Regular Army by Religion
and Nationality in 1848 and 1868
101

Table 3.1: Descriptive Statistics Across Years: Village/Urban Quarter-
Level Census Data (1897-1986)
130

Table 3.2: Descriptive Statistics Across Regions: Village/Urban Quarter-
Level Census Data (1897-1986)
131

Table 3.3: Descriptive Statistics: Individual-Level 1986 10 percent
Sample
132

Table 3.4: Impact of the Expansion of Modern Schooling on Inter-
Religious Literacy Gap (Evidence from Village/Urban Quarter- Level
Data)
133

Table 3.5: Impact of the Expansion of Modern Schooling on Inter-
Religious Literacy Gap (Evidence from Individual- Level 1986
Census Sample)
134

Table 4.1: Descriptive Statistics: Individual-Level 1848 and 1868 Census
Samples and 1897 Village/Urban Quarter-Level Full Census
172

Table 4.2: Descriptive Statistics: Medieval Arab Immigration Waves and
Churches and Monasteries of Egypt
173

Table 4.3: Coptic Population Share and Coptic-Muslim Economic Gap in
Nineteenth Century Egypt
175

Table 4.4: Coptic Regional Variation in Nineteenth Century Egypt and
Tax Revolts in the Eighth and Ninth Centuries
176

Table 4.5: Historical Origins of Coptic Spatial Distribution in Nineteenth
Century Egypt
177
x

Table 4.6: Which Historical Origins Are More Important? 178

Table 4.7: Determinants of the Spatial Distribution of Medieval Coptic
Institutions
179

Table C.1: Dimensions of HISCLASS 201

Table C.2: Occupational Titles in the Egyptian Census Records that Do
Not Exist in HISCO Scheme and Their Assigned HISCO Codes and
HISCLASS Values
202

xi

List of Figures

Figure 2.1: Occupational Distribution by Religious Group in 1848-68
Egypt (Traditional Sector)
90

Figure 2.2: State Industrialization and Christian-Muslim SSI Gap 91

Figure 2.3: Percentages of Christians in the Large "Overlapping"
Occupations in Traditional and Modern Sectors
92

Figure 3.1: Mean Literacy Rates of Muslims and Christians across
Decennial Birth Cohorts (1910-70)
124

Figure 3.2: Percentage of Muslims and Christians Working in White-
Collar Occupations across Decennial Birth Cohorts (1910-50)
124

Figure 3.3: Evolution of Supply of Elementary Schools (kuttabs) and
Modern Schools in the period (1907-69)
125

Figure 3.4: Regression of % Males Who Are Able to Read and Write
on % Non-Muslims (Village/Urban Quarter- Level Census Data 1897-
1986)
126

Figure 3.5: Regression of % Males Working in the Non-Agricultural
Sector on % Non-Muslims (Village/Urban Quarter- Level Data 1927-
1986)
127

Figure 3.6: Religion and Literacy across Birth Cohorts (1910-70)
(Individual- Level Data 1986 Census Sample)
128

Figure 3.7: Religion and Years of Schooling across Birth Cohorts
(1910-50) (Individual- Level Data 1986 Census Sample)
128

Figure 3.8: Religion and Employment in the Non-Agricultural Sector
across Birth Cohorts (1910-50) (Individual- Level Data 1986 Census
Sample)
129

Figure 3.9: Religion and Employment in White-Collar Occupations
across Birth Cohorts (1910-50) (Individual- Level Data 1986 Census
Sample)
130

xii

Figure 4.1: Social Status Index Distribution by Religious Group in
Rural and Urban Egypt (1848 and 1868 Pooled Individual-Level
Digitized Samples)
169

Figure 4.2: Wages and Poll Tax Burden in Medieval Egypt 170

Figure 4.3: Purchasing Power (in Kilograms of Wheat) of the Lowest
Bracket Annual Poll Tax, One Dinar (706-1450)
171

xiii

Abstract

This dissertation introduces economic theory, modern microeconometric methods,
and new microdata sources to the study of the economic history of the Middle East. In
particular, it examines an intriguing phenomenon in this field: the stylized fact that non-
Muslim minorities are better off, on average, than the Muslim majority in many countries
of the Middle East.
The Introductory Chapter provides an overview of the research questions and the
findings of the dissertation. It also describes the contribution of the dissertation to the
general economics literature as well as to the field of economic history.
Chapter 1 describes the digitization project of two nationally representative samples
of the 1848 and 1868 Egyptian censuses from the original manuscripts at the National
Archives of Egypt, and introduces an application of the data in the field of Middle
Eastern economic history. These censuses are perhaps the earliest in the Middle East and
among the earliest in any non-Western country to include information on all segments of
society on a wide range of demographic and socioeconomic variables.
Chapter 2 employs this new data source to examine the impact of state-led
modernization on the socioeconomic differentials between religious groups. Over the
nineteenth century, Egypt embarked on one of the world’s earliest state-led
modernization programs in production, education, and the army. I examine the impact of
this ambitious program on the long-standing occupational differentials and occupational
and educational segregation between Muslims, Christians, and Jews. The major finding is
xiv

that the modernization program did not attenuate the occupational gap or the occupational
and educational segregation structure between religious groups, because it did not focus
on providing access to skills to the unprivileged Muslim majority.
Chapter 3 examines the impact of the transformation of elementary religious schools
(kuttabs) into public modern primary schools in 1953 on the educational differentials
between religious groups in Egypt. Using several new data sources, the individual-level
census records from 1986, the village/urban quarter-level census records from 1897 to
1986, and the official schooling reports from 1907 to 1969, I find that the educational
reform reduced the literacy gap between religious groups by the end of the twentieth
century.
Chapter 4 attempts to answer the question: Why are local non-Muslims minorities of
the Middle East better off, on average, than the Muslim majority? It traces the origins of
the phenomenon in Egypt to the imposition of the poll tax on non-Muslims upon the
Islamic Conquest of the then-Coptic Christian Egypt in 640. The tax, which remained
until 1855, led to the conversion of poor Copts to Islam to avoid paying the tax, and to
the shrinking of Copts to a better off minority. Using new data sources that I digitized,
including the 1848 and 1868 census records, I provide both econometric and qualitative
evidence to support the hypothesis.
1

Introductory Chapter

This dissertation applies economic theory and quantitative methods to the study of
Middle Eastern economic history. The existing literature in the field of Middle Eastern
economic history has been traditionally limited to qualitative methods, and the more
recent use of quantitative methods was confined to the macro and institutional levels. In
contrast, the use of economic theory and quantitative methods in history, or the “New
Economic History,” has been fruitfully applied to the study of North America and
Western Europe. Applying these methods to the Middle East will allow a better
understanding of long-standing questions in the field, including the broad institutional
queries, as well as the posing of new questions.
For example, an intriguing phenomenon in the Middle East is that local non-Muslim
minorities are better off, on average, than the Muslim majority in many countries in the
region. This phenomenon has been known as a stylized fact in the literature (Courbage
and Fargues 1997; Issawi 1981; Kuran 2004; Tagher 1998; Krämer1989), but to date no
research provided comprehensive quantitative evidence on its existence or magnitude.
More importantly, no quantitative evidence has been provided on the evolution of this
phenomenon over the nineteenth and twentieth centuries, on its reaction to the
modernization process, or on its underlying historical origins. This dissertation introduces
new microdata sources including, most notably, individual-level samples of the 1848 and
1868 Egyptian census manuscripts that I digitized at the National Archives of Egypt, in
order to provide the first quantitative evidence on the phenomenon. These new data
2

reveal that among adult active men in mid-nineteenth century urban Egypt, 54 percent of
Christians, and 47 percent of Jews, worked in white-collar occupations, as opposed to
only 20 percent of Muslims. In rural Egypt, the figures stood at 21 percent for Christians
and 7 percent for Muslims (Jews were an urban population).
Using these new data and applying economic theory and econometric methods, the
dissertation tackles three new and inter-related research questions: What is the impact of
the nineteenth century state-led modernization in production, education, and the army in
Egypt on the socioeconomic differentials between religious groups? What is the impact
of the twentieth century educational modernization (the transformation from religious to
secular schooling) on these differentials? And why did these differentials emerge
historically in the first place? The dissertation focuses on Egypt, one of the largest
Middle Eastern countries, and one of the earliest countries in the region to adopt state-led
modernization in the first half of the nineteenth century as well as in the aftermath of
World War II.
The dissertation argues that the observed socioeconomic differentials between
religious groups in Egypt can be explained by a historical self-selection process across
religions, rather than by a causal impact of religion per se on economic outcomes. The
selection mechanism was induced by the Islamic poll tax institution that was introduced
at the time of the Islamic Conquest of Coptic Christian Egypt in 640, and remained until
1855. This tax led to the widespread conversion of poor Copts to Islam to avoid paying
the tax and to the shrinking of Copts to a better off minority. The persistence of the
differentials for over a millennium following the initial sorting process was due to the
3

guilds and elementary religious schools (kuttabs), institutions that allowed each religious
group to monopolize certain occupations by restricting entry of non-members.
The dissertation finds that a millennium later, state-led modernization in the
nineteenth century failed to reduce the differentials despite its westernized institutions of
production, higher education, and integration of the army, which provided alternative
upward mobility routes to access skills (job training, secular education, and military
training). State production projects merely recruited people based on their skills acquired
through guilds, and the other alternative routes to mobility were all very limited and did
not generate a general catching-up effect. Convergence in educational outcomes between
religious groups awaited the second major state-led modernization program, which took
place between1952 and1967. The educational modernization aspect of this program,
which transformed kuttabs into public modern primary schools and thus increased
Muslims’ access to skills, was more successful in closing the literacy gap by the end of
the twentieth century.
Chapter 1: A Pre-Modern Middle Eastern Population Brought to Light:
Digitization of the 1848 and 1868 Egyptian Individual-Level Census
Records

The recently discovered 1848 and 1868 Egyptian census manuscripts provide the
earliest microdata source on the socioeconomic differentials between religious groups in
Egypt. Carried out during the reigns of Muhammad Ali (1805-1848) and Ismail (1863-
1879) respectively, these censuses are perhaps the earliest in the Middle East and among
the earliest in any non-Western country to include information on all segments of society
including females, children, and slaves, and on a wide range of demographic and
4

socioeconomic variables, including religion, occupation, and school enrollment. Chapter
1 describes my digitization project of two nationally representative samples (about
80,000 records each) of the 1848 and 1868 Egyptian censuses from the original
manuscripts at the National Archives of Egypt, and introduces an application of the data
in the field of Middle Eastern economic history.
The digitized samples should be of interest to researchers in various fields: First, it
opens the door for Cliometricians to study the economic history of the Middle East, and
this dissertation is one example of the potential of this digitized data source in this regard.
Second, it provides historical demographers with an unbiased data source that represents
all segments of the society, and not just the wealthy strata. This allows the study of
demographic processes such as fertility, mortality, marriage, and migration. Finally, the
digitized samples are one of the rare microdata sources from the pre-colonial era, and
they would thus allow historians and social scientists to study of the impact of
colonization on the colonized populations.
Chapter 2: The Reluctant Transformation: Modernization, Religion,
and Human Capital in Nineteenth Century Egypt

Can state-led modernization act as a “melting-pot” for ethno-religious groups
reducing occupational differentials and occupational and educational segregation? The
economics literature has long examined the impact of segregation on economic outcomes
(Alesina and La Ferrara 2005). It is also known that segregation is endogenous and can
be altered by deliberate institutional interventions. State-led modernization (including
state industrialization) is one example of institutional interventions that may alter the
occupational and educational segregation structure in the labor market.
5

Egypt provides an interesting case study in this regard. Over the nineteenth century,
Egypt embarked on one of the world’s earliest state-led modernization programs in
production, education, and the army. Chapter 2 examines the impact of this ambitious
program on the long-standing occupational differentials and occupational and educational
segregation between Muslims, Christians, and Jews. I employ the digitized samples of the
1848 and 1868 Egyptian censuses to examine this question. I examine the impact of state
industrialization, defined as employment in a state production project, on religious
occupational differentials using the traditional production sector (by industry) in each
census year as a counterfactual. To evaluate the impact of educational and military
modernization on occupational outcomes by religious group, I use the variation across
birth cohorts in exposure to these “treatments.” Finally, I provide cross-tabulations and
segregation indexes to examine religious segregation in the modern institutions.
I find that the first wave of state industrialization widened the religious occupational
gap that was traditionally in favor of non-Muslims, but the second wave led to upward
occupational mobility among both Muslims and Christians, although it did not alter the
gap. Educational and military modernization, on the other hand, favored Muslims who
benefited from these institutions almost exclusively, but the impact was too limited to
induce a general catching-up effect. Overall, occupational and educational segregation
was not attenuated by modernization, both because the traditional institutions in
production and education were still the major routes for skill-acquisition, and because the
new routes for mobility that modernization created were themselves segregated.
6

The chapter contributes to several lines of the social sciences literature. It provides
perhaps the first econometric evidence on the impact of state-led modernization on the
integration (or lack thereof) of ethno-religious minorities in the labor market. It also
contributes to the literature on the impact of state industrialization on the skill level of the
labor force and how the strength of traditional labor market institutions can hamper the
efforts of the state (Gerschenkron 1962). A large body of literature examined the impact of
technological change on wage inequality via shifting the relative demand for skills
(Acemoglu 2002; Goldin and Katz 1998; Atack et al. 2004; Goldin and Sokoloff 1982). The
chapter shows that this impact may vary by religious group. Finally, it has implications for
the long-standing debate in sociology and political science on whether modernization
increases or reduces the possibility of ethno-religious conflicts (Durkheim 1984; Rogowski
1985; Newman 1991).
Within Middle Eastern Economic History, the chapter provides the first rigorous
quantitative analysis of two of the most important and intriguing questions in the field: (i)
Egypt’s “failed” state- led modernization in the nineteenth century and its impact on
guilds and other traditional institutions (Baer 1964; Owen 2002; Marsot 1984; Ghazaleh
1999; M. Fahmy 1954; K. Fahmy 1998; Heyworth-Dunne 1938), and (ii) the “privileged”
position of non-Muslim minorities in the Middle East in the nineteenth century
(Courbage and Fargues 1997; Tagher 1998; Issawi 1981). The findings shed light on the
historical reasons behind the persistence of the religious differentials in Egypt, despite
allegedly early attempts at modernization and integration of non-Muslims.

7

Chapter 3: From Kuttabs to Schools: Educational Modernization,
Religion, and Human Capital in Twentieth Century Egypt

Chapter 3 examines the impact of educational modernization, in particular, the
transformation of elementary religious schools (kuttabs) into public modern schools in
the mid-twentieth century, on the educational differentials between religious groups in
Egypt. Educational modernization was a major aspect of the Egyptian state-led
modernization program in 1952-67. Prior to this transformation, and over the second half
of the nineteenth century and the early twentieth century, non-Muslim minorities were
faster than Muslims to shift from religiously segregated kuttabs to modern private
schools. Muslims, on their part, relied on kuttabs and the limited supply of public modern
schools that were available. The chapter employs several new data sources in the
analysis: the individual-level census sample from 1986, the village/urban quarter-level
census data from 1897 to 1986, and the official schooling reports from 1907 to 1969. In
order to identify the impact of the transformation of schools in the individual-level data, I
exploit the variation across cohorts and provinces of birth in the local supply of modern
schools, and I allow this to vary by religious affiliation. The impact in the village/urban
quarter-level data is identified from the variation in the non-Muslim population share and
the lagged supply of modern schools across provinces.
I find that the complete transformation of kuttabs to free modern primary schools in
1953 led to the convergence of the religious gaps with respect to literacy by the end of
the twentieth century. I contrast this “melting-pot” modernization to an earlier wave of
expansion of modern, yet non-free, schooling in 1927-36. I find that the earlier non-free
wave did not reduce the literacy gap. The relative success of the 1950s’ program in
8

comparison to the nineteenth century program is due to its focus on providing access to
skills (through free primary schools) to the Muslim majority.
The chapter contributes to the literature on the impact of the expansion of the supply
of schools (Duflo 2001) and the shift from traditional to modern schooling (Yuchtman
2010) on improving economic outcomes. To the best of my knowledge, this is the first
research that extends this literature to examine the impact of the expansion of the local
supply of schools on the inter-religious educational differentials. It is also the first
research in Middle Eastern Economic History to provide quantitative and rigorous
analysis of Egypt’s highly controversial educational modernization program in the 1950s.
Chapter 4: On the Road to Heaven: Poll Tax, Religion, and Human
Capital in Medieval and Modern Egypt

But why are non-Muslims minorities of the Middle East, on average, better off than
the Muslim majority? Chapter 4 traces the origins of the phenomenon in Egypt to the
imposition of the poll tax on non-Muslims upon the Islamic Conquest of the then-Coptic
Christian Egypt in 640. The tax, which remained until 1855, led to the conversion of poor
Copts to Islam to avoid paying the tax, and to the shrinking of Copts to a better off
minority. Using new data sources that I digitized, including the 1848 and 1868 census
manuscripts, I provide econometric evidence to support the hypothesis. I find that the
spatial variation in poll tax enforcement, budget constraint, and tax elasticity of
conversion, measured by four historical factors, predicts the variation in the Coptic
population share in the nineteenth century, which is, in turn, inversely related to the
magnitude of the Coptic-Muslim gap, as predicted by the hypothesis.
9

The four historical factors employed in the analysis are: (i) the eighth and ninth
centuries tax revolts that were ignited by stricter tax enforcement, (2) the Arab (Muslim)
immigration waves to Egypt in the seventh to ninth centuries that Islamized the local
elites in the receiving districts. Local elites were responsible for the decentralized tax
collection, and, depending on their religion, exercised more (or less) lenience towards
Coptic taxpayers, (3) the Coptic churches and monasteries in the twelfth and fifteenth
centuries, which provided financial and physical aid to poor Copts, and (4) the route of
the Holy Family in their biblical flight to Egypt. While tax revolts and Arab immigration
capture the variation in tax enforcement, Coptic institutions and the route of the Holy
Family measure the variation in the attachment to Christianity or the tax elasticity of
conversion. Coptic institutions also have a direct effect on the extent of conversion
through relaxing the budget constraint for poor taxpayers.
The chapter contributes to the economics literature on the lasting impact of
institutional changes on economic outcomes (e.g. Acemoglu, Johnson, and Robinson
2001; Acemoglu et al. forthcoming). I show in this chapter that the Islamic Conquest of
Egypt that established the poll tax institution in 640 AD created unintentionally religious
human capital differentials that persisted for over a millennium. It is also related to the
literature on the impact of religion on economic outcomes (Botticinni and Eckstein 2002
and 2005; Barro and McCleary 2003 and 2005; Becker and Woessmann 2009; Boppart et
al. 2008; Boorooah and Iyer 2005; Chaudhary and Rubin 2009). I provide a new
explanation for the correlation between religion and economic outcomes based on the
historical self- selection into the religion, which is caused by factors unrelated to religion
10

per se (poll tax). Finally, the effect of the poll tax on the conversion of Copts to Islam and
the monopoly of Copts over administrative jobs in medieval and modern Egypt are both
old topics in the historical tradition (On poll tax and conversion: e.g. John of Nikiu 1916;
Severus 1907; Dennett 1950; Morimoto 1981. On Coptic scribes: e.g. Al-Maqrizi 2002;
Al-Muqqadasi 1877; Tagher 1998; Sheikho 1987; Samir 1996; Amer 2000). Yet, this
research chapter is, to the best of my knowledge, the first to argue that the poll tax led to
selected conversion of Copts to Islam (rather than systematic conversion), and that that,
in turn, generated the human capital gap. My key point here is shifting the attention from
identifying the reason(s) behind the Islamization of Egypt per se, a vast topic in the
historical literature that generated fierce debates, into identifying the reason(s) behind the
emergence of the Coptic-Muslim human capital gap. This gap is rarely emphasized in the
historical literature although the Coptic dominance of administrative occupations is rarely
disputed. The new data sources that I collected in support of the hypothesis may thus
shed a new light on both the Coptic-Muslim gap and the historical process of Islamization
of Egypt.

11

Chapter 1

A Pre-Modern Middle-Eastern Population Brought to Light:
Digitization of the 1848 and 1868 Egyptian Individual-Level
Census Records

1.1 Introduction
It is therefore necessary for Us to enumerate exactly the people of Our country so
that it may be a cause of its progress in civilization.

Muhammad Ali Pasha (1845)
1

The conducting of the census is a matter of benefit for the homeland, and whoever
understands its merit will strive body and soul for the sake of its execution.

Egyptian Census Order (1847)
2

In 1816, Muhammad Ali Pasha, the autonomous Ottoman viceroy of Egypt, initiated
an ambitious manufacturing program in what constituted one of the world’s earliest state-
led development experiments. The program, which failed in the end to transform Egypt
into an industrial nation, was so intriguing to historians that it remained one of the most
central topics in the social and economic history of modern Egypt and the Middle East at
large. Nevertheless, despite the vast historical literature on the topic (Al-Gritli 1952; M.
Fahmy 1954; Al-Hitta 1967; Marsot 1984; Owen 2002), little is known with respect to

1
Cuno and Reimer (1997), p. 213.
2
Cuno and Reimer (1997), p. 215.

12

the size of the workforce in these Egyptian manufactories, let alone the occupational,
ethnic, age, or religious distributions of the workers. Such detailed picture, which is
indispensable to evaluate the social impact of the manufacturing experiment on the
populace, was unattainable either because the data did not exist at all, or even if they did
exist they were too sporadic to form a complete picture of the topic.
In this context, the Egyptian individual-level census records from 1848 and 1868
offer two extraordinary data sources on the Egyptian population in this early period. The
two censuses, that were carried out during the reigns of Muhammad Ali (1805-48) and
Ismail (1863-79) respectively, are preserved in 6,592 handwritten registers at the
National Archives of Egypt (henceforth, NAE). They include information on a wide
range of variables including geographic location, name, ownership of dwelling, type of
dwelling, relationship to household head, age, gender, religion, ethnicity, nationality,
place of origin, occupation, and place of work, among other variables. The data are
recorded systematically for every individual in the household including females, children,
and slaves. By virtue of the wide range of information contained in the census records,
one is able to form a detailed picture of the social impact of Muhammad Ali’s state-led
development, as well as of many other central topics in Egyptian and Middle Eastern
social and economic history, such as education and slavery, that are currently not fully
understood because of the lack of data.
In an international perspective, the 1848 and 1868 Egyptian census records are
perhaps the oldest modern censuses in the Middle East and among the earliest from any
non-Western country. Several salient features make these two censuses “modern” and
13

distinguish them from the earlier Ottoman enumerations in the fifteenth and sixteenth
centuries or the other contemporary Ottoman censuses in the nineteenth century: First,
while the target population of the Ottoman enumerations was the taxpaying population or
the population eligible for military conscription (adult males), the Egyptian census
records enumerated all segments of the population including females, children, and
slaves. Second, in the same spirit of contemporary Western censuses, the Egyptian census
records kept standardized individual-level lists and not a mere “count” of heads (unlike
the earlier Ottoman enumerations). Third, the two Egyptian censuses recorded socio-
economic information (such as school enrollment and occupation for males) besides
demographic ones (such as location, gender, age, and place of origin). This shows an
interest on part of the Egyptian government in the “human capital” of the population.
The digitization of these individual-level census records is significant for researchers
in various fields. The census records open the door for Cliometricians to employ
quantitative methods in studying Egyptian and Middle Eastern economic and social
history at the micro- or individual-level. So far, the study of the history of this region has
been mostly based on qualitative methods, generally because of the lack of data. While
qualitative research is indispensable in understanding various historical phenomena in the
Middle East, the introduction of quantitative methods will definitely enrich the historical
research and will allow tackling a number of research questions that cannot otherwise be
addressed.
In addition to this, little is known of the demographic and socioeconomic
characteristics of the pre-modern populations in the Middle East. Such knowledge is
14

hindered by the lack of unbiased sources of information that can provide representative
snapshots of the population. The available sources (such as tax, cadastral, and court
registers) are usually confined to specific segments of the population (perhaps the
wealthier strata), and are hence of limited use for the historical demographer who is
interested in the overall characteristics of the population. By the inclusion of all segments
of the population, the census records will allow the study of the basic demographics of
this pre-modern Middle Eastern population: fertility, mortality, marriage, and migration,
as well as socioeconomic conditions.
Finally, a significant feature of the census records is that they come from the pre-
colonial period unlike comparable historical censuses from other developing countries
that were conducted under colonial administration (e.g. British India 1872). They hence
provide an early snapshot of a pre-modern and a pre-colonial population in its early
attempts towards modernization. When combined with later post-colonial censuses, this
source allows historians and social scientists to examine the impact of colonization on the
colonized population.
This chapter describes the digitization project of two nationally representative
samples of the 1848 and 1868 census records, which I undertook with the help of data
entry assistants at the NAE. The project, that was carried out over the period 2009-10,
resulted in the creation of two datasets from the 1848 and 1868 censuses each with about
80,000 individual observations. The two datasets are currently being integrated into the
historical censuses of the North Atlantic Population Project (NAPP) at Minnesota
Population Center (MPC) and will be disseminated on the web, with free access to all, as
15

soon as the integration is done. Upon their public dissemination, the two datasets will
hopefully boost quantitative research on the history of the Middle East.
The rest of the chapter is organized as follows. Section 1.2 describes the census
registers: the historical context in which the Egyptian censuses were undertaken, the
format of the census registers, and the enumeration methodology. Section 1.3 describes
the sampling strategy and the details of the digitization project. Section 1.4 discusses how
the digitized samples can shed light on Muhammad Ali’s manufacturing experiment, as
an example for the research questions that one can tackle using this new data source.
Finally, section 1.5 concludes.
1.2 The Census Registers: Historical Context, Format of the Registers,
and Enumeration Methodology

1.2.1 Historical Context

In 1845, Muhammad Ali Pasha ordered the undertaking of a nationwide census
enumerating every individual in the country (including females, children, and slaves).
The census operations started in the rural provinces as early as in 1846 but were then
extended to the urban ones in 1847 and were further repeated in 1848 in most of the rural
units that were already enumerated in 1846-7. Although a number of enumerations took
place in the 1850s and the early 1860s, these remained limited to scattered geographic
units and did not constitute a national census. Egypt had to wait for almost twenty years
to have its second wide-scale census in 1865-69 during the reign of Khedive Ismail
(1863-79). That second census is incomplete in the sense that it did not cover the entire
country. The result of these two census operations was the creation of standardized lists
16

of individual-level census records kept in 6,592 hand-written registers (the vast majority
of which belong to the years 1848 and 1868) that were since then preserved at the NAE.
Before the discovery of the 1848 and 1868 censuses in Egypt, the series of Egyptian
modern censuses was believed to have started in 1882 (a few months before the British
occupation). The 1882 census was widely known because it resulted in a published
census report at the village/urban quarter-level although the microdata did not survive.
3

The second published census was conducted in 1897, and was then followed by almost
regular decennial censuses (1907, 1917, 1927, 1937, 1947, 1960, 1966, 1976, 1986, 1996,
and 2006). However, the microdata for the censuses over the period 1882-1976 were
destroyed either by choice or by chance. This resulted in the anomalous situation that the
only available census microdata in Egypt are either the very early ones from 1848 and
1868 or the most recent ones (1986, 1996, and 2006).
Within the Middle East, the 1848 and 1868 censuses are perhaps the oldest censuses
to contain an enumeration of females, children, and slaves. On the one hand, the earlier
Ottoman tax registers (Tahrir Defterleri) in the fifteenth and sixteenth centuries
contained an enumeration of only the adult male population with very basic information
provided since the main concern of the government was the tax-paying population
(Barkan 1957; Cosgel 2004; Hütteroth and Abdulfattah 1977). On the other hand, while
the Egyptian censuses were the counterparts of the censuses undertaken in other parts of
the Ottoman Empire from the 1830s onwards (Cuno and Reimer 1997), two features of

3
There are a number of census registers at the NAE that date back to 1879-82. These registers include
individual-level census records as well as preliminary tabulations and might represent part of the largely
destroyed microdata of the 1882 census.
17

the Egyptian censuses make them distinct from the Ottoman ones: first, they contained an
enumeration of females and children, and second, they contained detailed information on
occupation of all males and school enrollment of male children. The contemporary
Ottoman censuses lacked these features. According to Karpat (1978), the first Ottoman
governmental census took place in certain parts of the empire in 1831-38, but was an
enumeration of households rather than individuals. The second census took place in 1844
and was an enumeration of the adult male population. Finally, the third census (1866-73)
that was limited to the Danube province was also an enumeration of adult males. In fact,
the first census in the Ottoman Empire that contained an enumeration of females took
place in 1881/82-1893 (Karpat 1978).
These comparisons naturally lead to the questions: Why did the Egyptian
government conduct these censuses? What were the motives behind them? Was the
Egyptian government attempting to imitate the experiences of European countries?
Answering such questions is in no way definite; nevertheless, it may be useful to
speculate on the answers. Indeed, one important incentive for conducting the 1848 and
1868 censuses was “controlling” the individuals. The census lists provided the basis for
tax collection, military conscription, and corvée work for public projects, just like earlier
and contemporary Ottoman enumerations, and they allowed the government to
effectively control internal migration between villages (which was not normally allowed).
Nevertheless, enumerating females and children and including socioeconomic variables
such as school enrollment cannot be explained by purposes of “controlling” the
population alone, and might have been motivated by other concerns about fertility and the
18

potential size of the labor force. Here, one may speculate, following Alleaume and
Fargues (1998), that the Egyptian students who were sent on educational missions to
Europe may have played a role in introducing the techniques of the European censuses
into Egypt. One may also speculate that the European technicians (mostly French and
Italian) whom Muhammad Ali and his successors employed in order to provide advice to
the administration may have also played a role. At any rate, the census orders suggest a
conscious attempt to imitate the experiences of the more advanced nations.
1.2.2 Format of the Census Registers
The census registers are organized by province, with each province having a number
of registers enumerating its population. In urban provinces (Cairo, Alexandria, and
Rosetta), the registers are further classified by district (which are in turn divided into
urban quarters or shiyakhas).
4
In rural provinces, the registers are classified into
provincial towns, villages, hamlets, and Arab tribes. Besides the aforementioned urban
and rural provinces, there are two single-city provinces that have one register each: Al-
Arish (in Sinai Peninsula) and Al-Qusayr (on the Red Sea coast).
The format of recording in urban provinces is on a dwelling-by-dwelling basis with
the heading of each page usually stating the urban quarter and the street name.
5
For each
dwelling, the dwelling type (e.g. ruined hut, tenement house) is recorded, followed by the
legal status of the dwelling (e.g. private ownership, religious endowment), the name of

4
Rosetta is treated as one single district and is hence directly divided into urban quarters.
5
See Appendix A for two scanned pages of the registers.
19

the property right holder, and the dwelling number in the street.
6
The households residing
in the dwelling are then recorded one by one. Each household starts with the word
“residence of” followed by the name of the household head, and the breaks between the
households are usually clearly marked. The household members are then recorded: males
followed by females, and within each gender free individuals are followed by servants
(who are also free) and then followed by slaves.
In rural provinces, the census records of the provincial towns, villages, and hamlets
are classified into: (i) locals, (ii) strangers, who are further divided into legal and illegal
immigrants,
7
and (iii) deserters, or those who escaped illegally from the geographic unit.
The locals are recorded according to the section (Hissa) of the geographic unit they are
residing in. The household of the Hissa’s headman (sheikh) is always recorded first
followed by the other households. Strangers and deserters, on the other hand, are
classified geographically according to their place of origin (for strangers), or place of
destination if known (for deserters). In large provincial towns, the census records are
classified by urban quarters and street names (if available), and are hence more similar to
the recording in the urban provinces. Finally, Arab tribes registers classify the members
of the tribe by the specific tribal group (gama’a) they belong to and/or by their location.
8

6
I found a governmental order that dates back to 1847 with respect to naming Cairo’s streets and
numbering its dwellings that is published in Sami (1928, Vol. 2, pp. 547-52). The order might have been
related to the ongoing census operations.
7
To control peasants’ flight, immigration from one village to another required a governmental permission
starting from 1829. People were “illegally” migrating from one village to another in order to avoid paying
the excessive taxes, the military conscription, and corvée in public works (Cuno 1992; K. Fahmy 1998).
8
For Arab tribes, the recording starts with the household of the headman of the tribe followed by the other
households in the tribe. The registers usually state the tribe’s name and its location.
20

No information is recorded on the dwelling’s legal status in rural provinces, perhaps
because private landownership in rural areas was not yet fully recognized by the state
(Cuno 1992).
For every individual in the census, a systematic list of information is usually
recorded that includes the following variables:
1. Name: Egyptian naming system uses the first name of the individual followed by
the name of the father, the paternal grandfather, and so on. The names in the records are
usually recorded up to the father’s name. Some individuals have their names recorded up
to the grandfather’s name. Females’ first names in urban provinces are usually not
recorded unless the female is the household head.
2. Relationship to household head: This is recorded in detail for most individuals.
Some confusion might occur in large rural households where the relationship recorded for
an individual might be relating the individual to the household head or to the preceding
household member.
3. Age: is recorded in full years (as of the next birthday) for every enumerated
individual. Age is categorical (with only two options: child and adult) for females in
urban provinces. Age for infants is sometimes recorded in months or even days.
4. Occupation: This is recorded mainly for males. The occupational titles are pretty
detailed and reflect the highly specialized guild system that was prevalent at the time. I
found about 3,700 distinct occupational titles in each census.
9
For male children, school
enrollment is recorded as an occupation. Unemployed and retired personnel are recorded

9
The occupational titles “scribe” and “scribe at the customs department” are counted as distinct.
21

as such. The name of the work establishment is usually recorded if the person is
employed in governmental enterprises. Occupation in rural localities is often missing
indicating the default occupation: farming. Similarly, for Arab tribes, occupation is often
missing indicating perhaps that they are shepherds.
5. Nationality: This includes two categories: inside or outside government’s control.
The concept of Egyptian nationality seems to have been quite developed in the censuses
as nationals of other regions of the Ottoman Empire are recorded as “outside
government’s control,” indicating that they are foreigners.
6. Ethnicity: This is usually missing for those inside government’s control who are
ethnically Egyptians. Sometimes categories such as “Son of an Arab” or “Son of the
country” are used to describe these individuals. For the individuals outside government’s
control, and for slaves (who are mostly inside the governments’ control), the ethnicity is
usually recorded (e.g. Turkish, Armenian).
7. Religion: This includes the following categories: Muslim, Christian, and Jew.
Often the denomination within Christianity and Judaism is recorded (e.g. Coptic
Christian, Rabbi Jew). Also, the denomination can often be inferred upon combining
religion with ethnicity (e.g. Armenian Christian).
8. Place of origin: It is recorded up to the village-level, but is not exactly the place of
birth. In 1848 census, children inherit the place of origin of their father. The wife does
22

not take the place of origin of her husband, and it seems that she takes the place of origin
of her father instead.
10

9. Infirmities: An infirmity is indicated if the individual has any (e.g. blind).
10. Marital Status: is also often recorded for females living with no adult males
present in the household.
However, despite the modern features of the Egyptian census records, an element of
trial-and-error is also evident. The standardized lists are often not so standard, with some
information omitted or mentioned irregularly. Information on religion and nationality are
often omitted so long as the “default” category is assumed (e.g. “Muslim” and “Inside
government’s control”). Confusion on part of the census takers over some fields can be
inferred. For instance, age, which is a categorical variable (child or adult) for females in
Cairo and Alexandria according to census instructions (Cuno and Reimer 1997), is
sometimes recorded in full years for females in these cities. In both censuses, there was
often some confusion on whether there should be an age restriction when recording the
occupation for males or not.
11
Spelling mistakes are numerous, and each scribe seems to
have, to some extent, his own way of recording information, some being keener on
recording details than others. The records also reflect the preliminary stage of the
statistical operations at the time. Apart from preliminary tabulations on the age and
occupational distributions at the street- or village section (Hissa)-level, the census

10
In the 1868 census, many children are recorded as from Cairo while their fathers are from another
province indicating that the concept of this variable was getting closer to the “place of birth.”
11
The occupation for males who are less than 6 years old is sometimes left blank. In other cases, the word
“child” or “infant” is recorded in place of the occupation.
23

operations in both years did not result in national statistical tabulations in published
census reports perhaps due to the unsophisticated techniques the census takers were
using.
1.2.3 Enumeration Methodology
Enumeration in the censuses seems to have largely followed the “de facto” principle.
This means enumerating individuals according to their actual, rather than permanent,
place of residence: A person who is not present in his permanent place of residence at the
time of enumeration is recorded in that place as “not enumerated”, and is enumerated
instead in his actual location.
12
Servants and slaves are recorded as part of the household
they are residing with. An exception to this principle appears in not enumerating the
individuals who are “protected” by foreign consulates. These individuals, although they
should have been enumerated according to the de facto principle, and who are mostly
(but not always) foreigners in the modern sense of the word, are recorded at the
household head level only with the note that they are to be enumerated by their foreign
consulates. Another important exception to this principle is military personnel (officers
and soldiers) and students in public schools (which were military boarding schools in
nature) who were recorded as not enumerated with the note that they are to be
enumerated in the census of the army. The members of their households, however, were
recorded and enumerated. Finally, Cairo’s notables in the census of 1848 (but fortunately

12
The “not enumerated” note is either mentioned explicitly in words or by assigning a “zero” beside the
name of the non-enumerated individual meaning that he/she is not added to the enumeration count of the
page. A person who is not enumerated usually has only his/her name recorded (and occupational title for
the military personnel). Age and other information are normally not mentioned for non-enumerated
individuals.
24

not in 1868) were not enumerated in their actual place of residence and were instead
recorded separately in a special register.
13
The information on these notables in 1848 was
scarce and only mentioned the household head’s name and the number of males and
females residing in his/her household. Notables seem to have been a broadly defined
group constituting about 6 percent of 1848 Cairo’s population (Cuno and Reimer 1997).
The 1848 census operations were explicitly mentioned in the census orders of 1845
and 1847 (Sami 1928; Cuno and Reimer 1997). Noticeably, there was no “census form”
and the census operations were conducted by the headmen of urban quarters and villages’
sections, under the supervision of an officer from the army. There were threats of harsh
punishments for the concealment of individuals. Each enumerated individual was issued a
“ticket” indicating that he/she has been enumerated. For the 1868 census, the only order
that I found so far is very brief and refers to some decree issued by the parliament (Sami
1928). However, the 1868 census operations seem to have been very similar to those of
1848 as far as one can infer from the format of the census registers.
1.3 Digitization Project and Sampling Strategy
The digitization project of the census records consisted of the creation of a 1 percent
sample of each of the 1848 and 1868 census records (5 percent in the two major cities,
Cairo and Alexandria). I undertook this project over the period 2009-10 with the help of
data entry assistants at the NAE. The result was the creation of two representative

13
According to the census order (1847) (Cuno and Reimer 1997), each quarter’s headman had to send to
the Prefect a list of the notables in his urban quarter. The Prefect was to request counts of males and
females in the household from the household head, which were to be recorded in the notables’ register. This
way, the census takers and the quarters’ headmen in charge of the census operations could not intervene in
the enumeration of the notables or access their houses. Similarly, the Khedivial Diwan was responsible for
collecting information on foreigners from the consulates.
25

samples (about 80,000 individuals each) from the years 1848 and 1868, which are
currently being integrated into the NAPP datasets and will be publicly disseminated once
the integration is complete.
In what follows, I will first discuss the target population and the sampling frame.
Second, I will discuss the sampling mechanism in detail. Third, I will compare the
sampling strategy to the U.S. 1850 census sample constructed by IPUMS, and to the
unreleased and uncompleted sample of the 1848 Egyptian census described in Alleaume
and Fargues (1998) and Fargues (2003).
1.3.1 Target Population and Sampling Frame
I define the target population for each of the two samples as the population that was
residing in Egypt at some point (the census night) in 1848 or 1868. The sampling frame,
however, can be different and I define it as all the available census records in the
surviving registers from both years. The discrepancy between the two concepts is due to
two problems: First, the missing registers (non-coverage problem), and, second, the
multiple enumerations of some geographic units over the periods 1846-48 and 1865-69
(duplicates problem). In this subsection, I will address each of these two problems in
turn. I want to emphasize, however, that the solutions to the two problems discussed in
this subsection are the “pre-sampling” solutions that have to do with re-defining the
target population and/or the sampling frame. Post-sampling solutions (i.e. creating
sample weights) are discussed in section 1.3.2.
26

1. Missing Registers:
14
Some census registers are missing. They might have been
destroyed or are undiscovered yet. With the exception of the cities of Damietta, Suez, and
Al-Qusayr (in 1868), which are entirely missing, the surviving urban provinces have
almost complete registers in both years. In order to evaluate the completeness of the
surviving registers of the rural provinces in the 1848 and 1868 censuses and the extent to
which the surviving census records represent the population of rural Egypt in these two
years, I use the first published census of 1882 as the reference point. The 1882 census has
a complete list of the geographic units in Egypt to which one can compare the available
list from 1848 and 1868. Table 1.1 shows this comparison at the province-, district-, and
village-levels. At the province-level, all the 1882 rural provinces are represented in 1848.
Within the “surviving” provinces, Al-Gharbiya and Isna have the most serious missing
registers problem at the village-level (coverage rate is 7 percent and 14 percent
respectively). In 1868, however, the problem seems to be more severe, since 6 out of 14
rural provinces are entirely missing, and 3 provinces (Al-Minya, Girga, and Al-Sharqiya)
have extremely low coverage rates. At this point I cannot confirm whether the 1868
census was a nationwide census whose registers did not survive or rather a partial one
that was not completed for some unknown reason.
Although the aforementioned comparison gives an idea of the degree of coverage up
to the village-level within each province, it does not tell how complete the individual
records are within each province. After all, there might have been missing individual

14
The problem of missing pages within a surviving register is also possible. However, it is generally very
limited in both censuses. In such cases, I assumed that the missing pages are random and I oversampled
from the surviving pages.
27

census records due to under-enumeration. To have an approximate idea on the extent of
completeness of the individual-level records in each province (both urban and rural), I
again use the 1882 census as the reference point under the assumption that under-
enumeration was less of a problem in that year. In particular, in order to find an estimate
of the true population size of each province, I calculate the “true” population of each
province in 1848 and 1868 based on its share of the total population in 1882 census (I
will discuss the validity of the assumptions underlying this calculation in section 1.3.2). I
then compare this “estimated” population size to the size of the population that has been
actually enumerated in the census registers; the latter I obtained by summing up the
population counts from all the geographic units in a given province.
15
In order to do this
comparison, a priori information on the size of the entire population in Egypt in both
years is required. For 1848, I used the widely accepted figure of 4,476,439 (Alleaume and
Fargues 1998). For 1868, I assumed a constant annual growth rate over the period (1848-
82) in order to calculate the population in 1868 (knowing the population in 1848 and
1882). Table 1.2 shows this comparison for the two censuses. Overall, the results are
similar to the coverage rates in table 1.1, although the enumeration counts in a number of
provinces actually exceed the estimated populations especially in 1848. Nevertheless,
these provinces have relatively high coverage rates in table 1.1.
The problem of missing registers results in discrepancy between the target
population and the sampling frame. I chose to treat this problem in the pre-sampling

15
Logistic problems at the NAE did not allow me to collect the enumeration counts for the 1868 census. I
instead estimated the actually enumerated population in each province in 1868 using the total number of
pages for each province (adjusted for blank pages and tabulations), and the average number of individuals
per page in the sample taken from that province.
28

phase as follows: If an entire province listed in the 1882 census is missing in the 1848 or
1868 census (or has only a very small number of its units represented in the registers),
16
I
re-defined the target population to exclude this province. If, on the other hand, a province
listed in the 1882 census has a significant number of surviving registers in the 1848 or
1868 census, I included its registers in the sampling frame, and oversampled from them.
Whether the sampling frame represents the target population in this province depends on
the validity of the assumption that the missing registers are randomly distributed in that
given province. In the post-sampling phase, I addressed this point by adjusting the
sampling weight (Subsection 1.3.2).
2. Multiple Enumerations: The enumerations for the 1848 census started as early as
in 1846, but were repeated in most of the geographic units in 1847 and 1848 perhaps in
order to raise the accuracy of enumeration and to mitigate under-enumeration of
individuals. Similarly for the 1868 census, there were early enumerations in 1865-67, and
enumeration operations seem to have been extended to 1869 in some provinces. In order
to keep the consistency between the target population and the sampling frame, and to
avoid double counting of individuals, only the register(s) from the latest year of each
geographic unit was included in the sampling frame. For instance, if a village was
enumerated in 1846, 1847, and 1848, I included only the register(s) from 1848 in the
sampling frame. This assumes that (i) the register(s) from the earlier year(s) enumerate

16
Besides the entirely missing provinces, I namely excluded Damietta in 1848, and Al-Minya and Girga in
1868. Damietta (by 1882 boundaries) had only one surviving district in 1848 that was actually enumerated
in that year as part of another neighboring province, Al-Daqahliya, and I thus sampled it as part of the latter
province. So Damietta was not sampled as an independent province. I decided to sample Al-Sharqiya in
1868 although its coverage rate is pretty low (1 percent at the village-level and 6 percent at the individual-
level) because it has 20 census registers, which is (arguably) a relatively large number.
29

exactly the same population of the unit, and, (ii) if there are more than one register from
the latest year that they enumerate different segments of the population and thus that they
should be added up. In specific provinces in the 1848 census, and due to logistic
problems, I discovered the existence of duplicate registers for the same unit after
sampling started, and hence very few units had higher chance of selection. To correct for
these duplicates, I adjusted their sampling weights in the post-sampling phase (Section
1.3.2). Table 1.3 shows the number of units with multiple enumerations (in different
years) in each province for each census. As is evident from the table, the problem arises
in very few provinces, and is more prevalent in the 1848 census.
1.3.2 Sampling Mechanism
1. Explicit Stratification by Province: I started sampling in each census by explicitly
stratifying the country into the provinces that are included in the sampling frame. In order
to obtain an estimate of the population size of each stratum, I used the relative share of
each province in 1882 census and the a priori information on the total population size of
the country in 1848 and 1868. This calculation is based on two assumptions: (i) that the
administrative boundaries between provinces did not change over the period, and (ii) that
the distribution of the population in 1848 and 1868 across provinces is the same as the
distribution in 1882. As for the first assumption, it suffices to notice that only 2.5 percent
of the geographic units of 1848 (less than 1 percent in 1868) belonged to a different
neighboring province in 1882. The second assumption may be justified for the
distribution across the rural provinces, but might be questionable for the urban-rural
distribution with presumably an increasing share of the urban provinces over time due to
30

urbanization and foreigners’ migration to Cairo and Alexandria in the second half of the
nineteenth century as well as the growth of the Suez Canal cities. However, table 1.2
showed that the actual enumeration counts in the urban provinces are not much different
from the estimated sizes of strata.
Setting the sampling rate at 5 percent in Cairo and Alexandria and at 1 percent in all
the other provinces, I calculated the targeted sample size in each stratum (province). Two
notes are in order: First, I chose to explicitly stratify the sample by province. By explicit
stratification I mean that I calculated the targeted sample size in each province, and I
sampled independently from within each province to achieve that sample size. The other
option that I did not choose, namely implicit stratification, was to apply the fixed
sampling rate of 5 percent on Cairo and Alexandria as one stratum, and 1 percent on all
the other provinces as another stratum, without the need for setting a targeted sample size
in each province.
17
In other words, under implicit stratification I would have taken one
page every 20 pages in Cairo and Alexandria and one page every 100 pages in all the
other provinces. My choice of explicit stratification by province is actually motivated by
the desire to make the geographic distribution of the sample approximately reflect the
true geographic distribution of the target population at the province-level. In particular,
given the missing registers problem, implicit stratification would have resulted in a
geographic distribution of the sample that reflects that of the surviving registers, which is
not of significance per se.

17
This is not to be confused with explicit stratification between Cairo and Alexandria on the one hand, and
the rural provinces on the other hand, which was essential under both scenarios discussed in the text since I
applied different sampling rates between these two groups.
31

Second, I calculated the provinces’ sizes using the 1882 census shares because I
could not find reliable a priori information on the 1848 or 1868 total population in each
province. The closest approximation is to use the sum of the enumeration counts from all
the available geographic units in each province. This, however, will result in the true
stratum size only if there is no missing registers problem or no under- (or over-)
enumeration. Since this is not always the case, it was perhaps safer to use the 1882 shares
in order to calculate the sizes of the strata. Table 1.4 shows the geographic distribution of
the two samples along with the targeted and actual sampling rates.
2. Sampling Rules: I applied systematic sampling by page within each stratum.
Given the targeted sample size, knowing the total number of pages of the registers of
each province, and by assuming an average number of individual records per page, I
calculated the total number of pages required in each stratum to reach the targeted sample
size and hence the interval of the systematic sample. I selected a page randomly in the
beginning of each stratum, and then I took the successive pages according to the
calculated interval. A selected page is accepted in the sample if at least 75 percent of it
has individual-level records. If a page is rejected, the closest preceding or succeeding
page, which satisfies the 75 percent rule, is taken on an alternating basis. If a page is to be
taken in the sample, I recorded all the households that start in the page along with
information on the dwelling they are residing in (including the total number of
households and individuals in the dwelling).
18
I also recorded the information on the

18
I defined the starting point of the household as the line including the information of the first member in
the household.
32

register code and page number to facilitate referring to the original register if need arises
in the future. In applying this rule, if a household starts in a previous page and continues
on the sample page it is not included in the sample. Similarly, if a household starts on the
sample page and continues in the following page I entered it in its entirety. This ensures
that all households have an equal chance of appearing in the sample regardless of their
size. Overall, systematic sampling was less costly than a pure random sample and was
much easier to apply by data entry assistants. It also ensures the geographic spread of the
sample within each stratum (implicit sub-stratification within the province).
A few notes on the sampling rules are in order. First, group quarters are dwellings
where individuals live together with no household relationships between them (e.g.
churches, monasteries, and jail). The individuals in these group quarters are recorded in
the registers as one single household. If the sample page has a group quarter starting in
the page, I entered the group quarter as a single household in its entirety. Otherwise, the
page is rejected (since it won’t satisfy the 75 percent rule) and a preceding or a successive
page is taken.
19
Second, fragments, which are the individuals enumerated separately from
their place of residence or their households for some reason, are taken into the sample if
they happen to be on the sample page. I entered the fragments as separate households in
the same way as they are recorded in the registers, with the note that they are fragments.
In 1848 Cairo, fragments are usually confined to the end of the urban district where the

19
The IPUMS 1850 U.S. census sample applies individual-level sampling to individuals residing in the
group quarters, instead of entering the group quarter in its entirety. The drawback of my approach is the
resulting increase in the standard error because the observations in the group quarter are more likely to be
correlated. Nevertheless, the very small number of group quarters that I found in the registers and the fact
that they are limited to Cairo and Alexandria mitigate such concerns.
33

census takers record individuals who were away at the time of the census, but came back
later on and were thus enumerated separately. In 1868 Cairo, however, fragments are
found within the pages of the regular enumeration. In other provinces from both years,
fragments are usually mentioned at the end of the geographic unit.
Finally, in a few registers in Alexandria in 1848, the households’ breaks within the
dwelling were not marked. In such dwellings, individuals were classified by nationality,
place of origin, and gender, regardless of the household they belong to. Household
relationships were sometimes mentioned (for males), but were mostly ignored for
females. For such cases, and since it was often impossible to identify the households
inside the dwelling, I decided to enter the dwelling in its entirety. In particular, if a page
is selected into the sample, all the dwellings that start on the page are taken into the
sample. Nevertheless, since most of the male household members do have common
nationality and place of origin (defined in section 1.3.2), it was mostly possible to
identify the male members of each household.
Most of the variables were coded either ex ante or during the data entry operations.
Through the data entry software, the data enterer was able to add new codes to the
variables as new categories are discovered in the registers. These coded variables are:
type of dwelling, legal status of dwelling, relationship to household head, title (both of
dwelling’s property right holder and of individual; e.g. pasha), legal status of the
individual, nationality (both of dwelling’s property right holder and of individual),
ethnicity (both of dwelling’s property right holder and of individual), religion, place of
origin (which I coded according to the 1882 census geographic coding), and the presence
34

of infirmities. On the other hand, numerical variables included number of households,
and number of individuals in the dwelling. Finally, the variables that were entered as text
are: register code, page number, street name, dwelling number, name of the dwelling’s
property right holder, age,
20
occupational title (both of dwelling’s property right holder
and of individual), individual’s name, and census taker’s (or data enterer’s) notes on both
the dwelling and the individual. After the data entry was complete, I coded the
occupational titles according to the Historical International Standard Classification of
Occupations (HISCO), which is the same coding scheme used in NAPP samples, up to
the most detailed five-digits classification. This will hence make the integration of the
Egyptian occupational titles to the NAPP samples easier. Examples of the individual
records in the digitized samples are shown in Appendix B.
3. Creating Post-sampling Weights: In the post-sampling phase, sampling weights
are required when calculating population means, totals, and proportions in order to adjust
for (i) different actual sampling rates across provinces (table 1.4), (ii) non-coverage of
some geographic units within the rural provinces (missing registers), and (iii) duplicate
registers for the same geographic unit (multiple enumerations) which resulted in a higher
chance of selection for some units. The final weight for each geographic unit is the
product of these three weights. In what follows I will discuss the weight adjustment
needed to correct for each of these three concerns.

20
Age is entered as text because it is categorical for females in Cairo and Alexandria. Also, for infants, the
age is recorded in days or months, so I entered it as text as mentioned in the register.
35

(i) Adjusting for Different Sampling Rates: Defining the probability of selection as
the total number of individuals in the sample divided by the total number of individuals
in the population, individuals in Cairo and Alexandria have a higher probability of
selection than the individuals in the rest of the provinces, and sampling weights must
adjust for that. Moreover, actual sampling operations may have resulted in different
actual sampling rates across provinces. I use the estimated population figures, based on
the 1882 relative population shares, as the denominator for the probability of selection in
each province. The sampling weight (W
A
) is defined as the reciprocal of the probability
of selection.
(ii) Adjusting for Non-Coverage: Missing registers within urban provinces is
negligible. I thus focus here on non-coverage within the rural provinces. I estimate the
probability of coverage of a geographic unit at the village-level in the rural provinces
using a Probit model.
21
The dependent variable of interest is a dummy for having a
surviving register in the relevant census. The regressors of interest are the characteristics
of the geographic units that are known for both missing and surviving units. These are:
location of the unit (I try three specifications with region, province, and district fixed
effects), type of geographic unit (village, hamlet or Arab tribe, and provincial town), and
population size. I estimated the regression for each year separately. The results of these
regressions are shown in table 1.5, and the sampling weight (W
B
) is defined as the
reciprocal of the estimated probability of coverage. From the table, one can see that larger
units have a higher chance of survival, perhaps because they have bigger registers that are

21
I am grateful to Ragui Assaad, University of Minnesota, for this advice.
36

less likely to be lost. Paradoxically, provincial towns have lower chance of survival than
villages, and hamlets and Arab tribes have a higher chance of survival than villages.
Finally, units in Upper Egypt are more likely to have a surviving register than units in
Lower Egypt.
(iii) Adjusting for Duplicates: For the units that have multiple surviving registers in
different years, only the register(s) from the latest enumeration was included in the
sampling frame, and hence no weight adjustment is needed for these units. However, in
very few cases, I discovered the existence of multiple enumerations only after sampling
started, and hence such units had higher probability of selection. The sampling weight for
these units has to be adjusted by multiplying the original sampling weight (W
A
*W
B
) by
the reciprocal of the number of duplicate enumerations that were included in the
sampling frame.
1.3.3 Comparison to the IPUMS 1850 U.S. Census Sample and Alleaume and
Fargues 1848 Egyptian Census Sample

My sampling strategy is similar to the methods used for the 1850 U.S. census sample
constructed by IPUMS (Ruggles and Menard 1995). The systematic sampling mechanism
was used for both samples. In the IPUMS sample, as in my sample, sampling is done
every given interval of pages. However, there are two main differences: First, while my
sampling unit is the page and the households within the page, the U.S. 1850 census
sample is a sample of dwellings. In particular, while the U.S. sample selects a random
line from the page and a dwelling is accepted if the line falls on the first individual in the
dwelling; in the Egyptian sample, and due to the irregular content of pages and the non-
numbered lines, the randomization is at the page-level. Once a page is selected, all the
37

households in the page are taken into the sample. Hence, clustering in the Egyptian
sample is at the page-level while in the U.S. sample is at the dwelling-level. Second,
stratification in the U.S. sample is implicit, since the 1 percent fixed sampling rate is
applied to all census records so as to choose 1 individual every 100 individuals. In the
Egyptian sample, the missing registers problem led to the choice of explicit stratification.
Alleaume and Fargues collected an unreleased and uncompleted sample of the 1848
Egyptian census, which is described in their 1998 article (henceforth, AF sample) and
Fargues (2003). The AF sample applies a two-stage stratified sampling strategy. Initially,
the country is stratified into 29 strata (Cairo’s 10 districts, Alexandria’s 5 districts,
Damietta, other urban governorates, and the 13 rural provinces).
22
In addition, rural
provinces were further stratified into three substrata: main town, rural localities (villages
and hamlets), and Arab tribes. In the first stage, a number of geographic units were
selected within each stratum based on an a priori sampling rate f
1
. If the stratum consists
of exactly one unit (such as a district of Cairo) it is selected with sampling rate 100
percent. Also, Arab tribes are treated differently and each tribe is taken with probability
100 percent. In the second stage, a number of households are selected within the selected
units with a sampling rate f
2
. If a household is chosen, it is entered in its entirety.
My sampling strategy is different from the AF sample in the following aspects. First,
in deciding the size of each stratum, the AF sample used cited population figures
available for the urban provinces in 1848, and calculated the population figure for the

22
Qena and Isna were a single province in 1848 (Ramzi 1994). Hence, there are 14 rural provinces in the
1882 census but only 13 in the 1848 census.
38

rural provinces (as one stratum) as the residual. I instead used the 1882 population shares
to estimate the population figure for each province. Second, by using systematic sampling
by page within each stratum (province), I applied implicit geographic sub-stratification
where the probability of selection of any geographic unit into the sample is directly
proportional to the number of pages of its register (in other words, its population). The
AF sample, on the other hand, applies explicit stratification within each rural province,
and oversamples from the central towns as compared to the rural localities. Third, the AF
sample treats the non-coverage problem by choosing close substitutes (in terms of size) to
replace the missing geographic units. The AF sample also chooses close substitutes for
the households if non-coverage is within the geographic unit (e.g. missing pages). I
instead use sampling weights to adjust for the non-coverage problem at the geographic
unit-level within a rural province. If non-coverage occurs within the geographic unit, I
assumed that the missing records are random and oversampled from the surviving records
in order to facilitate the sampling methodology.
1.4 Application: Exploring the Nineteenth Century Egyptian
Modernization Program

In this section, I first provide selected descriptive statistics from the 1848 and 1868
samples to give a general idea of the information contained in this data source. Second, I
discuss how this data source can provide useful insights as to evaluating the social impact
of the ambitious modernization programs carried out by Muhammad Ali Pasha, as an
example for the questions that can be answered using the digitized samples.

39

1.4.1 Descriptive Statistics
The highest frequencies of selected variables from the 1848 and 1868 samples are
shown in Table 1.6. I show the statistics for Cairo and Alexandria separately from all the
other provinces to emphasize the interesting contrast between urban and rural provinces.
Regarding the dwelling type, in both 1848 and 1868 about 73 percent of the individuals
in Cairo and Alexandria live in houses, unspecified, or unknown dwelling types, versus
100 percent in the other provinces. What is more interesting, however, is the
composition of the dwelling types in which the remaining 27 percent of Cairo and
Alexandria population resided in. About 16 percent of the population lived in low-status
dwellings. These mainly include ruined huts, courtyards, and single rooms. Moreover,
about 7 percent lived in multiple- household dwellings such as tenement houses. Finally,
3 percent lived in production sites, i.e. they resided in the place where they worked (e.g.
coffee shops and bathhouses).
Dwelling legal status is almost only found in Cairo and Alexandria. Private or public
ownership is the most dominant legal status as about 64 percent in 1848 and 71 percent in
1868 resided in privately or publicly owned dwellings. This is followed by Waqf
dwellings; i.e. dwellings owned by the pious or religious endowments of individuals or
entities (whether Muslim, Christian, or Jewish).
As for the individual-level variables, gender distribution seems quite balanced and
does not immediately indicate a gender bias in the sample. Muslims form the vast
majority in urban and rural provinces in both the 1848 and 1868 samples. Christians are
the largest religious minority followed by Jews. With respect to nationality, about 7
40

percent of the individuals are outside government’s control or foreigners in the urban
provinces, as compared to 1 percent in the rural provinces. Also, the percentage of slaves
is higher in urban provinces than in rural provinces, since the majority of the slave
population in Egypt (and in the Middle East at large) was involved in urban domestic
service. In 1868, however, there was a slight increase in the percentage of slaves in rural
provinces, perhaps due to the cotton boom that resulted from the American civil war
(1861-65), and that led to a significant improvement in the living conditions of the
Egyptian peasants and their financial ability to hold slaves. Finally, regarding ethnicity,
the vast majority of the individuals are locals. Turks represent the second largest ethnic
group in 1848, followed by blacks. The latter group, however, became the largest ethnic
minority in 1868.
1.4.2 Application
Historians have long debated various aspects of the intriguing manufacturing
program carried out by Muhammad Ali Pasha, who is widely regarded as the founder of
modern Egypt. Apart from the debate on the economic evaluation of the program with
respect to efficiency in production, which requires gathering data on the costs of
production and market prices, the census records can provide useful insight into the
social impact of the manufacturing experiment on the labor force. This social face of the
experiment is a central and an often ignored aspect of the manufacturing experiment. As
K. Fahmy (2009, p. 126) points out, it is important when evaluating Muhammad Ali’s
development experiment to take into account “the millions of Egyptian people whose lot,
as a direct result of what Mehmed Ali actually did achieve, was that of hardship and
41

suffering” (italics mine). In this subsection, I will discuss the type of information that one
can extract from the census records to shed light on this social side. I will focus on two
questions: First, to what extent did the manufacturing experiment by Muhammad Ali lead
to the disappearance of the traditional guilds? Second, did the factories established under
this experiment provide routes for upward mobility for the labor force? In discussing
these two questions, however, I want to emphasize that I am not trying to answer the
questions in this limited space. I am only giving examples to the sort of information that
the census samples can provide with respect to these questions.
The first question triggered a long debate among historians. M. Fahmy (1954) argued
that the manufacturing program led to the abolition of restrictions on labor that were
present under the guild system, and thus to the complete liberation of the labor market.
By contrast, Baer (1964) criticizes this argument and refers to the strong qualitative
evidence that shows that the guild system survived until the early twentieth century, and
that Muhammad Ali’s manufactories did not give the final blow to the medieval guilds,
although it might have indeed hurt specific guilds. In support of Baer’s thesis, Owen
(2002) points out that even the textile workers who were hurt most by the manufactories
did indeed manage to survive in the aftermath of the closure of many of the
manufactories. In the same direction, Marsot (1984) states that the guild system remained
largely unaffected by the experiment except in the occupations that were in direct
competition with the factories, although she also notices that the factory system opened
up possibilities for children to work and to be promoted “by training rather than by
heredity” in contrast to the guild rules (Marsot 1984, pp. 181-3). Nonetheless, a slightly
42

different viewpoint is provided by Ghazaleh (1999), who criticizes Baer’s thesis on the
grounds that although the guilds did indeed survive until early twentieth century they
were largely weakened by the manufacturing experiment and gradually came under state
control. In response to Baer’s claim that the artisanal guilds, which were essentially the
most affected by the manufacturing experiment, did not constitute more than one third of
the total number of guild members, Ghazaleh points to the difficulty of estimating the
total number of guild members and their breakdown.
The Egyptian censuses samples can provide useful insights into answering this
question. First, by having the occupational title recorded for every male, along with the
name of the work establishment if the person is employed by the governmental/public
sector, one is able to estimate the size of labor force in each of the manufacturing projects
(henceforth, modern sector) in 1848 under Muhammad Ali and twenty years later in 1868
after many of Muhammad Ali’s manufactories closed down. Second, equally important,
one is able to estimate the size and breakdown of the traditional guilds outside the
manufactories (henceforth, traditional sector). An important caveat here is that the census
records include information on occupations rather than guild membership. Although the
two variables are not the same, one may argue that the occupational title is a reasonable
proxy for guild membership, given the fact that there was a high degree of specialization
within the guild system, with “the extreme splitting of occupations into guilds of
specialized branches” (Baer 1964, p. 25). At any rate, holding this caveat in mind, the
census records allow us to estimate the extent to which the occupations/guilds actually
disappeared due to the emergence of the modern projects, and also, to determine the
43

specific occupations/guilds that were mostly hurt. Table 1.7 shows the descriptive
statistics from 1848 and 1868 on the size of the labor force (adult males) that was
employed in the modern sector. From the table one can see that the modern projects
employed in 1848 about 11 percent of Cairo’s adult male population (at least 15 years
old), but the percentage went down to 3 percent in 1868. In Alexandria, the second
largest city, the percentage of workers in the modern sector was about 2 percent in 1848
and went up slightly to 4 percent in 1868. The other provinces had very low share of the
population working in the modern sector in both years but the share was higher in 1848.
Overall, it seems that the modern projects in 1848 were not large enough to actually lead
to the complete disappearance of the traditional occupations/guilds. Moreover, the
percentage of workers employed in these projects fell sharply in 1868 as compared to
1848, with the corresponding increase in the relative size of the traditional
occupations/guilds. I have to emphasize, however, that Ghazaleh’s argument that the
guilds were weakened by state control even if they survived cannot be answered using the
census records.
Furthermore, one can extract more detailed information that can actually help
determine which guilds were hurt most by modernization. Table 1.8 shows the
breakdown (or occupational distribution) of workers in the modern and traditional sectors
in both years into ten aggregate occupational categories (same as HISCO one-digit
categories). This aggregate breakdown is for presentation purposes only, and the detailed
occupational categories given by the five-digits HISCO coding are available in the
digitized samples. From the table, one can see that 26 percent of the workers in the
44

traditional sector are working in the category “construction workers, carpenters, building
painters, transportation workers, and laborers.” About 20 percent of the workers in the
traditional sector are in the service occupations “cooks, servants, slaves, policemen,
military personnel, and assistants.” The occupational distribution in 1868 is quite similar.
Going through the detailed occupational distribution in both years, one can decide the
exact occupations/guilds that were affected most by modernization in both years.
The second question has to do with the routes of occupational mobility that were
provided by the modern sector in 1848 and 1868. While most of the historians admitted
with varying degrees that the workers in the manufactories suffered from conscription
into the factories and the bad working conditions they faced there (M. Fahmy 1954;
Marsot 1984; Owen 2002), little is known about the occupational distribution of the
workers in these manufactories, and how it compares to that in the traditional sector. This
knowledge is crucial to evaluate the degree to which jobs in the modern sector compare
(whether favorably or not) to those in the traditional sector, and hence, whether they
provided routes for upward or downward occupational mobility. To the best of my
knowledge, only Ghazaleh (1999) provides a list of workers broken down by occupation,
their working days, and their salaries in Khurunfish factory, the oldest textile
manufactory in Egypt that was constructed in 1816. While the censuses samples do not
have information on wages, they do provide detailed information on the occupation and
the work establishment (factory) of each male worker in the modern sector. Table 1.8
shows the occupational distribution of workers in the modern sector in 1848 and 1868 in
Cairo and Alexandria (the two cities with the vast majority of modern projects), and how
45

it compares to that in the traditional sector. The vast majority of the adult male workers in
the modern sector in 1848, about 69 percent, are in the occupational category
“construction workers, carpenters, building painters, transportation workers, and
laborers.” About 9 percent are working as specialized artisans “wood treaters, gunpowder
makers, textile workers, millers, and food processors.” The white-collar workers (scribes,
stores clerks, etc) constitute only 9 percent of the workers. In 1868, however, the
distribution seemed to have shifted more towards white-collar workers (17 percent) and
away from “construction workers, carpenters, building painters, transportation workers,
and laborers” (57 percent). This may have been the case because of the nature of the
modern transportation projects in 1868 such as railways, telegraph, and steam navigation,
which required more administrative/clerical jobs than Muhammad Ali’s manufactories.
This may suggest that the modern projects were actually a route for downward rather than
upward mobility for Egyptians in 1848, but might have provided a route for upward
mobility to white-collar jobs in 1868.
1.5 Conclusion
The 1848 and 1868 newly digitized census samples provide a rich source of
information on the Egyptian population in the nineteenth century, which can benefit
researchers in various disciplines. First, historical demographers can use the Egyptian
census samples to examine patterns of fertility, mortality, marriage, and immigration in
this population. Multigenerational households can also be studied using this source.
Moreover, given the availability of contemporary historical census records from other
countries in the NAPP database, international demographic comparisons will be feasible.
46

Second, Cliometricians and historians of Egypt and the Middle East can benefit from
these digitized samples in studying slavery, modernization, and Waqfs, among other
topics. Quantitative statements about the magnitude of each of these phenomena will be
feasible. Third, genealogists can make use of these census records in studying history of
families in Egypt. Fourth, the spatial data on location and addresses in the cities can
benefit researchers in spatial sciences and urban history of Middle Eastern cities in
reconstructing a detailed historical map of nineteenth century Cairo and Alexandria.
Overall, the 1848 and 1868 census records open up entirely new possibilities for
quantitative research in studying the social and economic history of Egypt and the Middle
East.

47

Table 1.1: Evaluating Coverage in the Rural Provinces
in 1848 and 1868
1882 Province Number of
Districts in
1882
%
Covered
1848
%
Covered
1868
Number
of Units
in 1882
%
Covered
1848
%
Covered
1868
Lower Egypt:
Al-Daqahliya 6 100 100 443 74 32
Al-Sharqiya 6 100 33 433 74 1
Al-Qalyubiya 3 100 0 163 84 0
Al-Gharbiya 10 80 0 547 7 0
Al-Minufiya 5 100 0 331 62 0
Al-Buhayra 6 100 0 304 48 0
Upper Egypt:
Al-Giza 4 100 100 167 93 90
Bani Suwayf 3 100 100 168 83 31
Al-Fayyum 4 100 50 91 97 91
Al-Minya 4 100 25 267 89 1
Asyut 10 100 100 322 55 57
Girga 5 100 20 190 67 1
Qina 4 100 0 109 62 0
Isna 6 33 0 107 14 0
Provinces are defined according to the 1882 census administrative borders and division. The percentages
of the 1882 districts and villages for which there exists at least one surviving register in 1848/1868
censuses are recorded in columns 3, 4, 6, and 7.

48

Table 1.2: Estimating Individual-Level Coverage Rate in Each
Province in 1848 and 1868

Province 1848 1868
Estimated
Population
Actual
Population
Coverage
Rate (%)
Estimated
Population
Actual
Population
Coverage
Rate (%)
Cairo 255,978 237,000 93 320,360 258,898 81
Alexandria 158,021 119,788 76 197,766 154,876 78
Rosetta 13,233 18,187 137 16,562 2,257 14
Al-Arish 2,005 2,311 115 2,509 1,820 73
Al-Qusayr 1,659 3,024 182 - - -
Al-Daqahliya 397,720 336,402 85 497,752 144,489 29
Al-Sharqiya 259,766 279,918 108 325,100 20,384 6
Al-Qalyubiya 173,804 178,232 103 - - -
Al-Gharbiya 627,764 42,287 7 - - -
Al-Minufiya 439,745 126,287 29 - - -
Al-Buhayra 270,526 144,583 53 - - -
Al-Giza 187,672 217,513 116 234,874 257,038 109
Bani Suwayf 133,099 190,176 143 166,576 44,424 27
Al-Fayyum 145,444 173,931 120 182,025 121,448 67
Al-Minya 201,507 238,457 118 - - -
Asyut 390,438 247,104 63 488,638 424,512 87
Girga 350,928 192,465 55 - - -
Qena and Isna 413,234 219,320 53 - - -
Provinces in this table are defined according to: (i) the 1882 census borders and division for the
geographic units that were found in the 1882 list, and (ii) the 1848/68 censuses borders for the units
that were not found in the 1882 list. Estimated population of each province (which is an estimate of
the true population size) is computed based on the province's share in 1882 population census.
Actual population of each province is the total enumeration count from the census registers. In 1868,
I estimated the actual counts based on the total number of pages of the registers of each province and
the average number of individuals per page. Coverage rate is the ratio of actual population to
estimated population.

49

Table 1.3: Duplicates: Geographic Units that Were Enumerated
More than Once by Province
Province Duplicates in 1848 Duplicates in 1868
Cairo 0 0
Alexandria 0 0
Rosetta 0 0
Al-Arish 0 0
Al-Qusayr 0 -
Al-Daqahliya 185 0
Al-Sharqiya 118 0
Al-Qalyubiya 0 -
Al-Gharbiya 15 -
Al-Minufiya 7 -
Al-Buhayra 2 -
Al-Giza 1 2
Bani Suwayf 1 9
Al-Fayyum 1 0
Al-Minya 9 -
Asyut 0 21
Girga 0 -
Qena and Isna 0 -
Provinces in this table are defined according to: (i) the 1882 census borders and division for the
geographic units that were found in the 1882 list, and (ii) the 1848/68 censuses borders for the
units that were not found in the 1882 list. A duplicate is defined as the geographic unit that has
more than one register belonging to different years over the periods 1846-48 (for the 1848 census)
and 1865-69 (for the 1868 census). Under this definition, a unit that has more than one register that
all belong to the same year is not a duplicate.
50

Table 1.4: Sampling in 1848 and 1868: Geographic
Distribution by Province, and Targeted and Actual Sampling
Rates

1848/1868
Province
1848 1868
Target
Pop.
Actual
Sample
Size
Target
Sample
Rate
Actual
Sample
Rate
Target
Pop.
Actual
Sample
Size
Target
Sample
Rate
Actual
Sample
Rate
Cairo 255,978 20,635 0.05 0.08 320,360 33,285 0.05 0.10
Alexandria 158,021 16,061 0.05 0.10 197,766 23,617 0.05 0.12
Rosetta 13,233 448 0.01 0.03 16,562 513 0.01 0.03
Al-Arish 2,005 51 0.01 0.03 2,509 26 0.01 0.01
Al-Qusayr 1,659 175 0.01 0.11 - - - -
Al-Daqahliya 397,720 6,374 0.01 0.02 497,752 5,039 0.01 0.01
Al-Sharqiya 259,766 3,012 0.01 0.01 325,100 3,257 0.01 0.01
Al-Qalyubiya 173,804 3,908 0.01 0.02 - - - -
Al-Gharbiya 627,764 7,369 0.01 0.01 - - - -
Al-Minufiya 439,745 5,661 0.01 0.01 - - - -
Al-Buhayra 270,526 3,135 0.01 0.01 - - - -
Al-Giza 187,672 3,509 0.01 0.02 234,874 2,590 0.01 0.01
Bani Suwayf 133,099 1,456 0.01 0.01 166,576 1,568 0.01 0.01
Al-Fayyum 145,444 1,489 0.01 0.01 182,025 2,403 0.01 0.01
Al-Minya 201,507 2,264 0.01 0.01 - - - -
Asyut 390,438 4,309 0.01 0.01 488,638 6,117 0.01 0.01
Girga 350,928 3,540 0.01 0.01 - - - -
Qena & Isna 413,234 4,212 0.01 0.01 - - - -
Provinces are defined according to the 1848/68 censuses administrative borders and division.
Target population is calculated based on the province's population share in the 1882 census.
Actual sample size is the number of individuals that were actually included in the sample in
each province according to the province's definition given above. Target sampling rate is the a
priori planned sampling rate. Actual sampling rate is the ratio of the actual sample size to the
target population in each province.
51

Table 1.5: Probit Model for Probability of Coverage in Rural
Provinces in the 1848 and 1868 Censuses
1848 1868
Dependent Variable: Dummy variable for having at least one surviving
register in the 1848/1868 census
log(population) 0.138***
(0.021)
0.314***
(0.025)
0.353***
(0.029)
0.046*
(0.027)
0.144***
(0.040)
0.166***
(0.049)
Provincial Town -0.132
(0.165)
-0.160
(0.195)
-0.222
(0.219)
-0.184
(0.196)
-0.029
(0.289)
0.115
(0.347)
Hamlet or Arab
Tribe
1.261***
(0.143)
1.652***
(0.172)
- 1.319***
(0.201)
0.974***
(0.219)
-
Lower Egypt -0.400***
(0.044)
- - -1.037***
(0.052)
- -
Province Fixed
Effects
No Yes No No Yes No
District Fixed
Effects
No No Yes No No Yes
N 3,959 3,959 3,336 3,671 2,115 1,440
Pseudo R
2
0.046 0.297 0.283 0.139 0.424 0.408
Log Likelihood -2,475.95 -1,823.27 -1,547.17 -1,516.80 -765.83 -575.57
LR Chi-Squared 236.36
(p=0.000)
1541.74
(p=0.000)
1218.45
(p=0.000)
487.94
(p=0.000)
1126.29
(p=0.000)
791.89
(p=0.000)
Standard errors are in parentheses. The observations are the combined list of geographical units of the
rural provinces in the 1848 (or 1868) census and the 1882 census. Dependent variable takes one if the
geographic unit has at least one surviving register in 1848/1868 census and zero otherwise. Regressors
include: (i) log(population) measured (a) using the 1882 census population shares for the units that
existed in the 1882 census, and (b) using the census registers actual enumeration counts for the units of
1848/1868 censuses that were not matched in the 1882 census. (ii) Type of the geographic unit with
"village" being the base category. The other two types are provincial town and hamlet or Arab tribe.
The type is measured using the recorded type in the 1882 list for the units that existed in this list, or
using the 1848/68 census types for the units that did not exist in the 1882 list. (iii) Lower Egypt region,
province, or district fixed effects. The regression is estimated for each year separately.

52

Table 1.6: Selected Descriptive Statistics from the 1848 and 1868
Samples

1848 1868
Cairo and
Alexandria
Other
Provinces
Cairo and
Alexandria
Other
Provinces
Dwelling Type

House/unspecified/unknown 73% 100% 73% 100%
Low-Status dwellings 16% 0% 17% 0%
Multiple-Household dwellings 7% 0% 6% 0%
Production sites 3% 0% 3% 0%
Dwelling Legal Status
Unspecified 25% 99% 15% 100%
Ownership (public or private) 64% 1% 71% 0%
Waqf (religious endowment) 11% 0% 14% 0%
Gender
Male 49% 50% 51% 50%
Religion
Muslim 89% 92% 86% 91%
Christian 6% 6% 4% 7%
Jew 1% 0% 1% 0%
Unspecified 4% 2% 9% 2%
Nationality
Outside government control 7% 1% 7% 1%
Legal Status
Slave 2% 0% 3% 1%
Ethnicity
Local 87% 97% 87% 97%
Turkish 3% 1% 2% 0%
Black 1% 0% 5% 2%
European 1% 0% 1% 0%
Levantine 1% 0% 1% 0%
Nubian 1% 0% 1% 0%
Sample Size 36,509 43,519 56,902 21,513

53

Table 1.7: Sectoral Distribution of Adult Active Male Workers in
1848 and 1868 Egypt

1848 1868
Cairo Alexandria Other
Provinces
Cairo Alexandria Other
Provinces
% Adult
males in
the
modern
sector
9.92 2.2 1.64 2.79 3.61 0.32
N 6,086 4,735 9,034 9,986 6,446 4,018
Sample is restricted to males who are at least 15 years old with non-missing occupational title.

54

Table 1.8: Occupational Distribution of Workers in the Modern
Sector in 1848 and 1868

1848 1868
Modern Traditional Modern Traditional
Engineers, physicians, pharmacists, ships'
masters
1% 1% 4% 1%
Judges, agents, teachers, religious
workers, artists
0% 5% 0% 6%
Administrative and managerial workers 2% 0% 2% 0%
Scribes, financiers, stores clerks, customs
clerks, post clerks
11% 5% 20% 5%
Merchants, street sellers, auctioneers,
water porters, slaves traders
1% 8% 0% 13%
Cooks, servants, slaves, policemen,
military, assistants
2% 15% 8% 18%
Farmers, animal husbandry workers,
fishermen
0% 35% 0% 14%
Wood treaters, gunpowder makers, textile
workers, millers, food processors
10% 11% 5% 14%
Shoe makers, blacksmiths, goldsmiths,
silversmiths, potters, stone cutters
5% 4% 6% 6%
Construction workers, building painters,
transportation workers, laborers
68% 17% 55% 22%
N (Males at least 15 years old with non-
missing occupational title)
856 18,999 525 19,925

55

Chapter 2

The Reluctant Transformation: Modernization, Religion, and
Human Capital in Nineteenth Century Egypt

2.1 Introduction

Look at this battalion… There are there Arabs and Copts, Mussulmans and
Christians, that march in the same rank. I assure you that not one of them
troubles himself about his comrade’s religion. Equality between them is complete.

Ismail, Khedive of Egypt (1863-79)
1

Over the nineteenth century, Egypt, at the time an autonomous Ottoman province,
embarked on one of the world’s earliest state-led modernization programs in production,
education, and the army, preceding the Japanese program by half a century. Even though
the program failed to transform Egypt into a developed economy,
2
the creation of
westernized institutions triggered a social transformation. In 1848, state production
projects employed 8 percent of the adult active male population of urban Egypt, and in
1868 3 percent.
3
By the end of the century, the railways became the largest employer in
the country (Toledano 1998, p. 261), and workers in large state transportation enterprises

1
Charmes, Gabriel (1883), Five Months at Cairo and in Lower Egypt, p. 161.
2
The average annual growth rate of real GDP per capita in Egypt was standing at a low level of 0.4 percent
over the period (1820-70), in contrast to 1.1 percent in the U.S. and Western Europe, or even to 0.7 percent
in the non-industrialized neighboring Lebanon over the same period (Pamuk 2006).
3
Author’s calculations from the Egyptian 1848 and 1868 digitized census samples (See section 2.3).
56

became the pioneers of the workers’ collective action movement in the early twentieth
century.
4
The newly established army (which grew to be 5 percent of Egypt’s adult active
male population in 1868) later on played a pivotal role in the nationalist movement, and
the graduates of the modern schools (despite serving only 1 percent of males 5-25 years
of age in urban Egypt) were the pioneers of Egyptian “enlightenment.”
5

Did religious groups benefit equally from industrial, educational, and military
modernization? Did modern institutions integrate traditionally occupationally-segregated
religious groups as Khedive Ismail believed? Following a long medieval tradition, non-
Muslims in Egypt were relatively more concentrated than Muslims in white-collar
occupations such as scribes, moneychangers, and merchants (See figure 2.1). They also
had better educational outcomes in terms of quantity and quality.
6
In chapter 4, I trace the
origins of these religious differentials to the imposition of the Islamic poll-tax on non-
Muslims upon the Islamic conquest of the then-Coptic Christian Egypt in 640 AD, which
led to the widespread conversion of poor and uneducated Copts to Islam to avoid paying
the tax, and to the shrinking of Copts into a minority that is wealthier and more educated,
on average, than the growing Muslim majority. Traditional institutions of skill-
acquisition (the guilds and the kuttabs or elementary religious schools) then preserved

4
Examples include the strikes by Cairo tramways’ workers in 1908 and 1911 and the railways’ workers in
1908, and their role in the first workers union in Egypt in 1909 (Beinin and Lockman 1987, pp. 57-82).
5
See K. Fahmy (1998, pp. 263-8) for an insightful discussion of the “unintended” impact of the move to
create a regular army on building the national sentiments in Egypt. Heyworth-Dunne (1938, pp. 159-63,
177-80, 253-64, 269-71, 304-7, 326-9) provides biographies of the prominent graduates of the schools.
6
School enrollment rate in kuttabs among male children 5-14 years of age in 1848 Cairo was 34 percent for
Muslims, 51 percent for Christians, and 80 percent for Jews. See section 2.2.2 for a discussion of the
quality differences between Muslim, Christian, and Jewish kuttabs, and section 2.3 for the data source.
57

religious differentials and segregation for centuries. Noncompetitive policies in these
institutions constrained access to skills and made occupations hereditary within each
religious group (See section 2.2). Modern institutions can affect religious differentials via
two channels: First, the technological change embodied in state production projects
creates a new demand for labor that may differ in its complementarity to skills from the
traditional production sector. A large body of literature examined the impact of
technological change on wage inequality via shifting the relative demand for skills (e.g.
Acemoglu 2002; Goldin and Katz 1998; Atack et al. 2004; Goldin and Sokoloff 1982).
This impact may vary by religious group. Second, modern institutions create new routes
for skill-acquisition: job training in state production projects, education in modern public
schools, and military training in the army. Black and Lynch (1996) demonstrated that job
training has a positive impact on productivity, and Yuchtman (2010) found a positive
impact of educational modernization in nineteenth century China on economic outcomes.
Costa and Kahn (2006) found a positive effect of military experience on integrating
minorities in the labor market. The impact of these skill-acquisition routes on human
capital may also differ by religious group.
I evaluate the effect of modernization on religious occupational differentials using a
new and unique data source: individual-level samples of the Egyptian 1848 and 1868
censuses that I digitized from the original manuscript forms. I focus on three areas of
modernization: (i) the two waves of industrial modernization in 1816-48 and 1848-68, (ii)
the four public modern higher-education schools of engineering, medicine, veterinarian
medicine, and translation operating from 1827 to 1854, and (iii) the conscription of non-
58

Muslims as soldiers into the army starting from 1856 and the promotion of Egyptians into
commissioned officers’ ranks starting from 1854-63. I also examine whether religious
occupational/educational segregation in the modern institutions (state production
projects, modern schools, and the post-reforms army) was different from that in the
traditional institutions. I distinguish between two hypotheses: (1) what I call integrated
modernization, in which modernization acts as a “melting pot” reducing both religious
occupational/educational segregation and differentials, and (2) segregated modernization,
in which modernization fails to reduce religious segregation but the differentials can
either increase or decrease.
7
The data allow me to evaluate these hypotheses because they
include information on religion, occupation, school enrollment, and work
establishment/school, in addition to geographic and demographic information.
I examine the impact of state industrialization, defined as employment in a state
production project, on religious occupational differentials using the traditional production
sector (by industry) in each census year as a counterfactual.
8,9
I attempt to disentangle the
mechanisms of this impact, technological change’s relative demand for skills versus job

7
This distinction is inspired by the modernization literature in sociology and political science (Newman
1991). “Melting pot modernization theory” predicts that the role of religion subsides as societies modernize
and move to “organic” solidarity, which is based on division of labor rather than religion (Durkheim 1984,
pp. 126-46). “Conflictual modernization theory” suggests to the contrary that modernization invigorates
religious conflicts through the competition of previously segregated groups over the new economic
opportunities that modernization creates (Rogowski 1985). While it is not my goal in this chapter to
evaluate the impact of modernization on religious conflicts, examining the impact of modernization on
religious differentials and the occupational integration of religious groups has its implications for this
literature.
8
Endogeneity of religion is not a concern since my goal is to examine the changes in religious differentials
that were caused by modernization rather than to examine the levels of these differentials.
9
I use the 1848 census to evaluate the first industrial wave and the 1868 census to evaluate the second.
59

training. To evaluate the impact of educational and military modernization on
occupational outcomes by religious group, I use the variation across birth cohorts in
exposure to these “treatments.” Finally, I provide cross-tabulations and segregation
indexes to examine religious segregation in the modern institutions. In all the analysis, I
draw on historical evidence and rigorous quantitative methods to support the findings.
The economics literature has long examined the impact of ethno-religious
segregation on economic performance (e.g. Alesina and La Ferrara 2005). It is known in
this literature that segregation is endogenous and can be altered by deliberate institutional
changes. The chapter provides perhaps the first empirical evidence on the impact of state-
led modernization on the integration (or lack thereof) of minorities within the same
occupations/schools/military ranks in the labor market. The chapter also contributes to
the literature on the impact of state industrialization or the so-called “forced
industrialization” on the skill level of the labor force and how the strength of traditional
labor market institutions can hamper the efforts of the state (Gerschenkron 1962). It also
has implications for the long-standing debate in sociology and political science on
whether modernization increases or reduces the possibility of ethno-religious conflicts
(Durkheim 1984; Rogowski 1985; Newman 1991).
The chapter provides the first rigorous quantitative analysis of two of the most
important and intriguing questions in Middle Eastern economic history: (i) Egypt’s
“failed” state-led modernization in the nineteenth century and its impact on guilds and
other traditional institutions (Baer 1964; Owen 2002; Marsot 1984; Ghazaleh 1999; M.
Fahmy 1954; K. Fahmy 1998; Heyworth-Dunne 1938), and (ii) the “privileged” position
60

of non-Muslim minorities in the Middle East (Courbage and Fargues 1997; Tagher 1998;
Issawi 1981). The digitized census samples allow me to provide not only the first
quantitative evidence on these two phenomena, but also the first examination of the
impact of state-led modernization on the religious differentials, a question that received
less attention in the historical literature. The findings shed light on the historical reasons
behind the persistence of religious differentials in Egypt, despite allegedly early attempts
at modernization and integration of non-Muslims.
The rest of this chapter is organized as follows. Section 2.2 provides a historical
background. Section 2.3 describes the data. I introduce the empirical analysis and the
results on industrial modernization in section 2.4, and on educational and military
modernization in section 2.5. Finally, section 2.6 concludes.
2.2 Historical Background
2.2.1 Guilds and State Production Projects
At the turn of the nineteenth century, the Egyptian labor market was organized
through the guild system (ta’ifa) that encompassed almost all labor force in the cities
(Baer 1964, pp. 16-48). Apprenticeship, the sole route for skill-acquisition and accession
to mastership, and the jedik, a fee required to practice a craft in a workshop, both made
guild membership hereditary and thus restrained occupational mobility (Baer 1964, pp.
49-76, 107; Raymond 1973, pp. 544-51). Barriers to entry were higher in artisanal and
trade (high-skill) occupations than in transport and services (low-skill) occupations, since
the latter had neither a clear-cut apprenticeship (Baer 1964, p. 62) nor a jedik system.
Apprenticeship complemented kuttabs as a route for acquiring skills in white-collar
61

occupations (Heyworth-Dunne 1938, p. 87),
10
and there were significant barriers to entry
into such occupations even in the government.
11
With the growing state power in the
nineteenth century, the authority of the guilds headmen increased as they became
responsible for conflict resolution between the government and guilds members (Baer
1964, pp. 77-84).
Religious segregation of guilds reflected religious occupational segregation. Figure
2.1 shows the occupational distribution of the adult active male population of each
religious group in the traditional sector in 1848-68. Christians and Jews were relatively
more concentrated in white-collar and high-skilled guilds (scribes, carpenters, merchants,
jewelers, moneychangers, tailors).
12
A similar pattern of segregation existed in the late
eighteenth and early nineteenth centuries. Up to the nineteenth century, there was no
religious segregation within occupations, and “mixed” occupations, i.e. those that had
sizeable memberships from different religions, had single guilds (Raymond 1973, pp.
524-6).
In 1816, Muhammad Ali (1805-48), the Ottoman viceroy of Egypt, launched the first
wave of state industrialization by establishing the first textiles manufactory in Egypt

10
Dor Bey, the Swiss education inspector under Ismail (1863-79), pointed out that “Coptic (Egyptian
Christian) children have acquired a skill in arithmetic through practical exercises when accompanying their
fathers to government offices” (Tagher 1998, p. 213).
11
Lord Cromer, the British consul-general of Egypt (1883-1907), observed that Copts limited access to
white-collar occupations, such as scribes and accountants, by making the accounting system “archaic” and
“incomprehensible to anyone but themselves.” They resisted all attempts at reform or simplification of the
system. The duke of Harcourt, a French contemporary author, emphasized the peculiarity of the Coptic
accounting system and how it allowed Copts to dominate administrative and fiscal jobs for centuries
(Tagher 1998, pp. 212-3).
12
All the statistics come from the author’s calculations from the digitized census samples (See section 2.3).
62

(Owen 2002, p. 69). This was followed by a series of manufacturing projects throughout
his reign that, in 1848, included textiles (66 percent of Egypt’s modern sector
employment), military industries (17 percent), and other industries (17 percent), such as
printing, paper, coin making, and wood (Al-Gritli 1952, pp. 51-65; M. Fahmy 1954, pp.
21-54). 70 percent of these production projects were in Cairo and Alexandria,
13
and they
differed from traditional workshops in both technology and size. Unlike traditional
workshops, state projects employed machines that were a crude imitation (by Egyptian
blacksmiths under the supervision of European skilled artisans) of the technology used in
Western Europe at the time. A few manufactories used steam power to operate the
machines, but the vast majority resorted to animal power. The estimated median size of
the manufactory in 1848 sample is 163 workers, much larger than the size of a traditional
workshop.
To fund his ambitious projects, M. Ali monopolized internal and external trade, and
centralized the tax system (Owen 2002, pp. 65-6). Nonetheless, by 1868, 67 percent of
his manufactories closed down, and employment in the surviving ones in Cairo and
Alexandria dropped by 75 percent. Traditional accounts for the failure center around: (1)
the Anglo-Turkish commercial convention (1838) which abolished monopolies and
reduced tariffs in the Ottoman Empire, and (2) the London treaty (1841) which limited
the size of the Egyptian army, the raison d’être of the manufactories (Owen 2002, pp. 75-
6). Owen, however, suggests that there were structural causes for the failure such as the
fading centralized power of the state since 1837, the unqualified personnel, the fuel and

13
Author’s calculations, based on the production projects observed in the 1848 census sample.
63

power problems, and, perhaps most importantly, the failure to create an entrepreneurial
class.
M. Ali’s successors (1848-68) focused on transportation in the second wave of
industrial modernization. Projects such as the railways (1853), telegraph (1854), steam
navigation companies (1856 and 1863), and Alexandria tramways (1861) (Al-Hitta 1967,
pp. 215-91) recruited 58 percent of Egypt’s modern sector employment in 1868. The
share of military industries remained stable at 22 percent, while the share of textiles fell
sharply to 3 percent. Other industries, such as printing, tannery, and coin making
recruited 17 percent of modern sector workers. Cairo and Alexandria were still the major
centers of modernization (86 percent of the projects), although railways, steam
navigation, and telegraph served most of the country. The technology used in these
projects was presumably more sophisticated than the technology of the 1848
manufactories, but the median size of the plant dropped to 60 workers. Both internal and
external debts were used to fund these projects, and the increasing debt (especially for
Suez Canal which was opened in 1869) was the major reason behind Egypt’s fall under
British occupation in 1882. Generally, however, the transportation projects were more
successful than the earlier manufactories and many of these projects survived until today.
2.2.2 Kuttabs and Modern Schools
In 1800, religious institutions were the sole providers of education through the
kuttabs (Heyworth-Dunne 1938, pp. 2-7, 84-92). The curricula of these elementary
schools were mainly religious, but there were important differences between Muslim,
Coptic (Christian), and Jewish schools. Coptic schools taught arithmetic and geometry
64

besides religious subjects, perhaps to prepare the students for their future careers in the
administration,
14
a preparation that was supplemented later by apprenticeship.
15
Little
evidence exists on Jewish schools at the time, but it appears that Hebrew was “taught not
only for religious reasons but also for practical purposes” (Heyworth-Dunne 1938, p. 92).
Yet, Muslim kuttabs focused only on learning Arabic orthography through memorizing
the Quran. 20 percent of the teachers in Muslim kuttabs in 1848 were blind, and thus
were unable to teach reading and writing, unlike teachers in Coptic and Jewish kuttabs.
Higher education was provided solely through Muslim religious institutes and was thus
focused on religious preparation of the ulama.
16
Overall, the non-Muslim educational
system was more inclined towards “useful” knowledge than the Muslim system.
17

Public modern schools were introduced in 1816 in reverse order starting with higher
schools, and then followed by preparatory schools and primary schools. Four public
higher schools stand out as the most significant: medicine (1827-54), engineering (1834-
54), veterinarian medicine (1827-51), and translation (1836-51).
18
Non-Muslims were not

14
Heyworth-Dunne (1938, p. 85) mentions the following passage as the earliest account on Coptic schools
written by Sadlier (1693): “... the children were taught religion, good manners, to read and write Arabic
and Coptic... and were taught geometry and arithmetic because these two sciences are very useful and
necessary on account of the overflowing of the Nile, whereby the limits are lost; so that it becomes
necessary for them to measure out their land, and by the benefit of the first of these sciences they compute
the yearly increase.” Italics are mine to identify the words of Sadlier (1693).
15
See the discussion in section 2.2.1 and footnote 10.
16
Al-Azhar was the foremost renowned religious higher educational institution in Egypt since the tenth
century. The ulama are the Muslim clergy.
17
See Mokyr (2002 and 2005) on the concept of “useful” knowledge.
18
Although the schools of medicine and engineering were reopened in 1856 and 1858 respectively, they
were so frequently reopened and closed that they became generally in a bad condition until 1863, the year
of accession of Ismail to power (Heyworth-Dunne 1938, pp. 320-3). Therefore, it is reasonable to conclude
that the indicated periods of operation were indeed the effective ones until 1868.
65

admitted to these schools until 1873 (Sami 1928, p. 1123),
19
and thus they presumably
had no access to higher education until then (Heyworth-Dunne 1938, p. 87). Nonetheless,
most of the public modern schools were closed in the 1840s and 1850s, perhaps in
response to the London treaty (Heyworth-Dunne 1938, pp. 223-43, 288-301, 313-23).
2.2.3 National Regular Army and Military Reforms
In 1822, M. Ali ordered the conscription of Egyptian Muslims into the army for the
first time in centuries (K. Fahmy 1998, pp. 89-92). The measure was taken in order to
build a regular army, following the style of Napoleon Bonaparte’s revolutionary army
that conquered Egypt in 1798-1801, and to replace the Mamluk irregular military
regiments constituting mainly of slaves and mercenaries, which were inferior to European
armies in organization and tactics (K. Fahmy 1998, pp. 79-84). During Egyptian military
history in the nineteenth century, two military reforms are perhaps the most critical: First,
in 1856, non-Muslims were conscripted as soldiers for the first time; an action that may
have led to the integration of non-Muslims in the army (Tagher 1998, pp. 203-4).
20

Second, during Sa’id’s reign (1854-63), Egyptians were allowed promotion to high
commissioned officers’ ranks, which were previously preserved for the ruling Turkish

19
According to the 1867 law, non-Muslims were to be allowed to enter governmental kuttabs and primary
schools (Heyworth-Dunne 1938, p. 363).
20
This occurred in response to the abolition of the Islamic poll-tax on non-Muslims in 1855 (Tagher 1998,
p. 204). Conscription was carried out on an arbitrary basis, by the village headmen, and not through a
universal scheme (K. Fahmy 1998, pp. 97-9).
66

(Muslim) elite (Al-Raf’i 1987, p. 35). Non-Muslims were not officially excluded from
this latter reform.
21

2.3 Data
To examine the impact of state-led modernization, individual-level data with
information on religion, occupation, and work establishment are required. I digitized two
nationally representative samples (about 80,000 records each) of the 1848 and 1868
Egyptian censuses from the original Arabic manuscripts, which are preserved at the
National Archives of Egypt, and I also constructed an oversample of non-Muslims in
Cairo in both years.
22
The Egyptian censuses are perhaps the earliest individual-level
census records in the Middle East to include information on every member in the
household including females, children, and slaves. They include information on a wide
range of variables including location (province, district, quarter/village, street/section,
dwelling number), dwelling ownership, dwelling owner’s name, dwelling type, dwelling
size, name, household relationships, age, gender, ethnicity, nationality, religion, place of
origin (province, district, village), legal status (free or slave), occupation (for males),
school enrollment (for male children), enterprise/school (for active individuals), and

21
Although Heyworth-Dunne (1938, p. 338) reports that the Coptic Pope’s request to allow Copts to
become army officers was not met until the Pope’s death in 1861 when it was dropped, there are reasons to
conclude that non-Muslims were not banned from becoming officers, although they may well have been
subject to state discrimination. First, there is the factual evidence of observing two Coptic officers in the
1868 sample as opposed to the 1848 sample where there is none. Second, Dunn (2005, p. 25) points out that
Sa’id allowed both Copts and (Muslim) peasants to become officers. Third, there are no theoretical grounds
for banning non-Muslims from becoming officers since conscripting them was, by itself, a violation of
Islamic law. Finally, Ismail (1863-79) recruited foreign non-Muslim officers widely in the army, and it is
hard to believe that there was an official ban that applied only to local non-Muslims.
22
Sampling rates are 8-10 percent in Cairo and Alexandria and 1 percent in the provinces. The
oversampling of non-Muslims is at the rate of 25 percent. The data source and sampling strategy are
described in Chapter 1.
67

infirmities. This list exceeds that of the 1850 U.S. census,
23
but the latter records literacy
and real estate value in addition.
In this chapter, I restrict the sample to adult males who are at least 15 years old. Four
key variables in the empirical analysis require close attention: First, occupational titles
were first recorded in full text in Arabic as they appeared in the manuscripts. I then
manually coded the occupational titles (about 3,700 distinct titles in each census)
following the five-digit Historical International Standard Classification of Occupations
(HISCO) scheme, where I created new codes for the titles that were not found in HISCO.
Based on this occupational coding, I constructed my main occupational outcome
measure, Social Status Index (henceforth SSI), which classifies occupational titles into 12
ordered categories (from unskilled farm workers to higher managers). There are four
criteria used in this classification: manual vs. non-manual, supervisory vs. non-
supervisory, skill level (from low to high), and primary sector vs. non-primary.
24
I also
created an alternative outcome, Occupational Group, where I collapsed the 12 ordered
categories of the SSI into: white-collar workers, skilled workers, and unskilled workers.
Second, religion is recorded for most observations in 1848 but not in 1868, where I
inferred an individual’s religion from his name. The main religious groups are Muslims,
Christians, and Jews. The last two groups can be broken down further by denomination,
when combined with ethnicity. Christian denominations include Copts (Egyptian

23
The additional variables in the Egyptian censuses are detailed geographic location, dwelling ownership,
dwelling type, religion, detailed place of origin (internal migration can be identified), and enterprise/school.
24
See Appendix C and van Leeuwen and Maas (2005) for details.
68

Christians), Armenians, Levantines, and Ruum (Ottoman Greeks), while Jewish
denominations are Rabbanites and Karaites.
Third, the state industrialization measure (Modern) is an index that takes the value of
one if the individual is employed in a state production project. The census takers are
generally keen on distinguishing between being employed by the state (miri) and working
in the private sector (barrani). For most of the individuals employed by the state, the
establishment of work is recorded in the occupation field. I constructed a list of state
production projects, by project’s name, industry, and location in both 1848 and 1868,
based on information from secondary historical sources (M. Fahmy 1954; Al-Gritli 1952;
Sami 1928). I then combined the list with the digitized census samples in order to decide
whether the individual’s establishment of work is a state production project.
25

Finally, I created a categorical measure, “Industry,” to identify the individual’s
industrial affiliation. The categories include textiles, transportation, military, other
modernized industries, and non-modernized industries.
26
The latter are defined as the
industries that exist only in the traditional sector, while the first four categories exhaust
all industries in the modern sector.
Table 2.1 shows the descriptive statistics for the sample of adult males who are at
least 15 years, including those with missing age. Around 3 percent (5 percent) in urban
Egypt are “Without Occupation,” 3 percent (3 percent) are out of labor force (students

25
The criteria for the SSI, including primary sector vs. non-primary, are inferred from the occupational title
only. Information on work establishment that is used to construct the state industrialization index is not
used in constructing the SSI. Thus, a livestock worker is assigned the same SSI regardless of being
employed in a given “modern” state textiles manufactory or on a “traditional” farm.
26
An animal-driver and a scribe in the railways both belong to the “transportation” industry.
69

and retired), and 8 percent (12 percent) have missing occupational titles in 1848 (1868).
In rural provinces, the percentages of unemployed and out of labor force are negligible
but the percentage missing is particularly large at 27 percent (31 percent) in 1848 (1868).
About half of the adult active males in urban Egypt are unskilled workers, but the
percentage is much higher in the rural Egypt at more than 80 percent, since this category
includes farmers. The religious composition is quite stable across the two years and
across urban and rural provinces, with Muslims constituting the vast majority (about 90
percent), followed by Christians (6-7 percent), and Jews (1 percent), but the latter are
mainly urban.
27
Throughout the empirical analysis (sections 2.4 and 2.5), the sample is
restricted to active males who are at least 15 years old and who have non-missing values
for all the variables in the analysis.
2.4 Empirical Analysis: State Industrialization
This section examines the impact of state industrialization, i.e. being employed in a
state production project (Modern), on occupational attainment and whether this impact
varies by religious group. Since profit-maximization may not have been the underlying
incentive behind the program, the external validity of the results is restricted to other
experiments of state industrialization. For example, state projects deviated from the free
markets norm in their recruitment policies. While the 1868 wave recruited individuals
who applied voluntarily for the jobs, many workers were drafted into the 1848

27
The vast majority of Christians are Coptic Christians (86 percent). Non-Coptic Christians are
concentrated in Cairo and Alexandria where they constitute 40 percent of Christians. Almost all Christians
in rural Egypt are Copts (98 percent). Jews are urban, and in Cairo are either Rabbanite (56 percent),
Karaite (29 percent), or unspecified.
70

manufactories (Owen 2002, p. 76; Marsot 1984, pp. 181-5), and headmen of urban
districts carried out the drafting (Ghazaleh 1999, p. 122; Sami 1928, Vol. 2, p. 374).
Many workers were discontent and cases of flight were recorded (Ghazaleh 1999, pp.
122-3). Historical evidence suggests that matching skills with job requirements was the
main recruitment criterion, be it by drafting or voluntary application.
28

Conceptually, Modern represents the equilibrium in the labor markets of state
production projects. Thus, the marginal effect of state industrialization on occupational
attainment is a “combined” effect of labor supply and demand. Supply-side factors
include, besides observable individual characteristics, the unobserved individual skill that
may affect the decision to apply for or be assigned to a job at the state projects.
29

Demand-side factors include the technological change that dictate the job structure in the
state projects, the availability of job training as an alternative route to acquire skills away
from guilds and kuttabs, and the recruitment policies in the projects. These factors may
vary by religious group differentially across the modern and traditional sectors. I first
introduce the results on technology-skill complementarity of state industrialization and its
impact on religious occupational differentials. I then attempt to disentangle the two
mechanisms of industrialization’s impact: technological change’s demand for skills

28
A state order in Sami (1928) (Vol. 2, p. 279) asserts that recruits into the cotton spinning and weaving
manufactories have to be of those who had some knowledge of spinning and weaving. Similar examples are
found regarding recruiting turners, carpenters, and blacksmiths who were to imitate the imported European
machines (Owen 2002, p. 70), railways engineers (Sami 1928, Vol. 3, p. 242), tarboush (hats) weavers and
dyers (Vol. 3, p. 24), shipbuilding engineers (Vol. 3, p. 330), and telegraphers (Vol. 3, p. 459).
29
Skills were constrained by the guild structure. Jews, despite having higher school enrollment than
Christians in 1848 Cairo, did not work as scribes, which was a Christian-dominated guild. They instead
dominated the financial jobs. I will discuss the recruitment policies of state modern projects in 2.4.4.
71

versus job training. I next address the threats to the empirical strategy. I finally compare
religious occupational segregation in the modern and traditional sectors.
2.4.1 Technology-Skill Complementarity of State Production Projects
Did the technology employed in each industrial wave increase the relative demand
for skills, compared to the traditional sector, or was it low-skill-biased? Did technology-
skill complementarity vary by industry? Evidence from the U.S. suggests that the First
Industrial Revolution was “de-skilling,” (Goldin and Sokoloff 1982; Atack et al. 2004),
but that the relative demand for skills rose with the Second Industrial Revolution (Goldin
and Katz 1998). I thus estimate the following ordered categorical regressions:
1

= =

+

+

+

2

= =

+ !"

+ # !"

×

+

+

Where the probability that the SSI of an individual i, residing in district j, takes the
value m (m = 1, …, 12) is a function that depends on Modern in (1), and on both
Industry dummies, as defined in section 2.3, and a full set of interactions between
Modern and Industry in (2). is a vector of individual characteristics: age, slave
dummy, foreigner dummy, black dummy, and migrant dummy. are district fixed
effects to account for variation in spatial proximity to the state projects.
30
is an error
term. Notice that Industry exhausts all the industries that exist in the modern sector.

30
The limited number of districts in 1848 and 1868 (97 and 53 respectively) relative to the number of
observations (18,146 and 18,968) mitigates the concerns about the incidental parameters problem when
including district fixed effects in the ordered categorical regression.
72

Factor changes in odds ratios are shown in table 2.2. Panel (A) shows that
recruitment in a state modern project reduces an individual’s odds of having the highest
SSI value versus the combined lower values in 1848 (compared to the odds for an
individual in the traditional sector), but improves his odds in 1868. For example, being in
the modern sector in 1848 reduces the probability of working as a clerk (SSI = 9) by 0.05,
but increases the probability of working as an unskilled manual worker (SSI = 2) by 0.20.
On the other hand, in 1868, modernization increases the first probability by 0.08, but
decreases the second probability by 0.05. Thus, the technology of the 1848 wave was
low-skill-biased, i.e. it increased the relative demand for low-skilled workers, while the
technology of the 1868 wave was complementary to high skills. The 1848 regional
estimates suggest a stronger low-skill-bias in rural provinces than in urban provinces.
Panel (B) shows that the effect of state industrialization varies by industry; modern
textiles that dominated the 1848 wave is low-skill-biased, while modern transportation
that dominated the 1868 wave is high-skill-biased. For example, being in a modern textile
manufactory in 1848 compared to traditional textiles increases the probability of being an
unskilled manual worker by 0.55 [95 percent CI = (0.47, 0.63)]. But working in a modern
transportation enterprise compared to traditional transportation in 1868 increases the
probability of being a clerk by 0.29 [95 percent CI = (0.22, 0.33)].
31
The results are
qualitatively similar across urban and rural provinces. The findings seem to be consistent

31
Traditional textile workers are primarily skilled, such as weavers and spinners, while traditional
transportation workers are primarily unskilled, such as animal drivers and boatmen. Throughout section
2.4, all CIs are calculated by Delta method at the 95 percent confidence level, and all average marginal
effects on the probability of a particular outcome are for a free, non-black, non-migrant, and non-foreigner
individual.
73

with the evidence from industrialized countries. Goldin and Sokoloff (1982) showed that
textiles made an intensive use of women and children, as unskilled workers, in the early
U.S. industrialization. Also, improvements in transportation increased the relative
demand for skills in the U.S. in the early twentieth century (Goldin and Katz 1995).
2.4.2 State Industrialization, Religion, and Occupational Differentials
Did religious groups benefit (or were hurt) equally by the new demand for labor that
the projects created or were specific groups poised to benefit more than others? Did state
industrialization reduce or aggravate religious occupational differentials? Did this impact
vary by industry? I estimate the following equations to tackle this question:
3

&''!(
)

= =

+

+
#*ℎ
"

+ ,-.

+

× *ℎ
"

+ /

+

4

= =

+ *ℎ
"

+ #-.

+
!"

+
/
!"

× *ℎ
"

+ 1

+ 2
!"

×

+ ,

× *ℎ
"

+ 3
!"

×

×
*ℎ
"

+ 4

+

Where Occupational Attainment = SSI or Occupational Group. The probability
distribution of each outcome in (3) depends, besides Modern, on Christian and Jew,
indexes for religious affiliation, and the interaction of Christian and Modern. I do not
include an interaction of Jew and Modern because of the small number of Jews in the
modern sector. In (4), MainIndustry = textiles in 1848 and transportation in 1868
74

Table 2.3 shows the factor changes in odds ratios from the ordered logit regressions
of equations (3) and (4) where SSI is the dependent variable. The odds of having the
highest SSI value versus the combined lower values for a Christian or a Jew in the
traditional sector are greater than for a Muslim. The findings of table 2.2 hold for the
Muslim majority; working in the modern sector implies downward mobility among
Muslims in 1848 but upward mobility in 1868. Christians in the modern sector, however,
experienced upward mobility in both waves. Thus, for a Muslim in the modern sector, the
probability of working as a clerk is lower by 0.06 [CI = (-0.07, -0.05)] in 1848 but higher
by 0.06 in 1868 [CI = (0.01, 0.11)], compared to a Muslim in the traditional sector.
However, for Christians in the modern sector, the effects are 0.12 [CI = (0.07, 0.16)] in
1848 and 0.15 [CI = (0.09, 0.20)] in 1868. Overall, modernization widens the religious
average SSI gap in 1848, but does not alter it in 1868 [See figure 2.2 for the effect of
industrialization on the religious differential in probability of each SSI value].
32

Moreover, the impact varies by industry; working in modern textiles in 1848 caused
downward mobility among Muslims but modern transportation in 1868 improved their
lot. Christians, however, enjoyed upward mobility in both industries in 1848 and 1868.
As an alternative specification, I use Occupational Group as the dependent variable
in (3), and I estimate a multinomial logit regression. The rationale is to relax the
proportional odds assumption and to allow the coefficients to vary by occupational
outcome. The estimated average partial effects are in table 2.4. Non-Muslims’ advantage

32
A wider religious average SSI gap means that the ratio of Christians’ odds of having the highest SSI
value versus the combined lower values compared to Muslims’ odds in the modern sector is larger than the
corresponding ratio of Christians’ to Muslims’ odds in the traditional sector.
75

in the traditional sector in 1848 and 1868 is in both skilled and white-collar occupations.
Recruitment in the modern sector reduces a Muslim’s probability of being a white-collar
worker by 0.05 [CI = (-0.07, -0.03)], but increases it by 0.20 [CI = (0.14, 0.26)] in 1868,
compared to a Muslim in the traditional sector. For Christians in the modern sector, the
effects are 0.29 [CI = (0.16, 0.43)] in 1848 and 0.34 [CI = (0.13, 0.56)] in 1868. The
religious white-collar gap is wider in the modern sector in 1848 but not in 1868.
In general, while industrialization widened the Christian-Muslim occupational gap in
1848, it did not affect it in 1868. This stems from the downward mobility of Muslims in
the 1848 modern textile manufactories, and their upward mobility in the 1868 modern
transportation enterprises. Christians, on the contrary, experienced upward mobility by
employment in state modern projects in both industrial modernization waves and in both
textiles and transportation. In the next subsection, I will examine the mechanisms that
may account for these findings.
2.4.3 Mechanisms of State Industrialization: Technological Change or Job
Training?

Two mechanisms may account for the findings in tables 2.3 and 2.4: (i)
Technological Change: In 1848, the technological change embodied in the modern sector
decreases the relative demand for Muslims’ skills, but increases the relative demand for
Christians’ skills, compared to the relative demand for skills of their coreligionists in the
traditional sector. In 1868, technological change increases the relative demand for both
Muslims’ and Christians’ skills. Under this mechanism, guilds and kuttabs still dominate
the acquisition of skills, and state production projects are merely hiring workers based on
76

their skills.
33
(ii) Job Training: State production projects provide job training for skill-
acquisition. Training effect can be either positive or negative depending on the relative
average SSI of the “training-based” jobs in the modern sector compared to that of the
jobs for which there is no job training provided. According to this mechanism there is
positive training effect for Christians in both 1848 and 1868. The training effect for
Muslims is negative in 1848 but positive in 1868.
To examine these mechanisms, I classify occupations in the modern sector into two
types: “overlapping” or “guild-based” occupations, i.e. those that exist in both traditional
and modern sectors; and “new” or “training-based” occupations, i.e. those that exist
solely in the modern sector and are thus presumably created by modernization.
34
In this
classification, the “overlapping” occupations, which recruited 73 percent and 66 percent
of modern sector employment in 1848 and 1868 respectively, proxy for the technology
effect, because they were controlled by guilds in the traditional sector that provided a
priori training through apprenticeships. “New” occupations, on their part, are a proxy for
the training effect, because they did not possess guilds to provide training to employees

33
The observation that the three Jews who are observed in the modern sector in the sample are all in
modern banks may also be explained using the technological change and the relative demand for skills
mechanism. Since skills are guild-specific, Jews did not traditionally possess a skill in administrative
occupations, and their comparative advantage lied in financial occupations. Hence, it was only the
technology of the modern banks that benefited Jews by increasing the relative demand for their skills.
34
“New” occupations include, for example, telegraphers, train drivers, ticket conductors on trains and
trams, factory workers, and printing workers. Almost every occupation that exists in the traditional sector
has a guild, and so it is safe to assume that the “overlapping” occupations are “guild-based.” A few
“overlapping” occupations, such as engineers, physicians, and military officers are education-based or
military-based rather than guild-based, and are thus an outcome of modernization in its broader sense. Also
job training might have been provided for some of the “overlapping” occupations. I will examine the
impact of modernization on professional and military occupations in section 2.5. However, since my
interest here lies in identifying the training effect of industrial modernization per se, I chose a rather
conservative definition of “new” occupations that can be safely attributed to job training provided by these
projects.
77

and so had to resort to job training, which is supported by historical evidence.
35
I
construct an index for working in a “new” occupation, “New,” and I extend equation (3)
to be:
5

= =

+

+ #*ℎ
"

+ ,-.

+
4

× 6.

+

× *ℎ
"

+ 3

× 6.

×
*ℎ
"

+ /

+

Notice that the technology effect for Muslims in the modern sector is captured by
Modern, while the training effect for Muslims is captured by the interaction of Modern
and New. For Christians, the effects are augmented by Christian. The factor changes in
odds ratios are shown in table 2.5. In 1848, the technology effect for a Muslim in the
modern sector reduces his odds of having the highest SSI by a factor of 0.43. The training
effect reduces the odds for a Muslim in a “new” occupation of having the highest SSI by
an additional factor of 0.07 compared to a Muslim in an “overlapping” occupation in the
modern sector. For Christians the two effects operate in opposite directions but the
technology effect dominates; there is a positive technology effect of 2.85 (0.43*6.63), but
a negative additional training effect of 0.004 (0.07*0.05). In 1868, the technology effect
for both Muslims and Christians is positive but is stronger for the latter; it increases a

35
Evidence includes workers in glass manufacturing (Sami 1928, Vol.2, p. 368 and 376), sugar refining
(Vol. 2, p. 376), cotton spinning and weaving (Vol. 2, p. 384), silk spinning and weaving (Vol. 2, p. 421),
telegraph (Vol. 3, p. 240), machine building (Vol. 3, p. 331), printing (Vol. 3, p. 659), and gun making
(Vol. 2, p. 384 and Vol. 3, p. 992). An 1866 state order (Vol. 3, p. 652) introduced telegraph as one of the
subjects to be taught to students in public schools to satisfy the need for telegraphers in the railways.
78

Muslim’s (Christian’s) odds of having the highest SSI by a factor of 1.91 (5.39). The
additional training effect is insignificant for Muslims but is negative for Christians.
The results of table 2.3 are thus mainly explained by technological change.
Christians, who traditionally had a skill advantage over Muslims, gained more from state
industrialization that increased the relative demand for their skills in both 1848 and 1868.
Muslims were hurt in 1848 because of the low-skill-bias of the textiles technology, but
benefited in 1868 because of the high-skill-bias of the transportation technology. The
effect of job training in the new occupations, although perhaps more open to individuals
than guild-based occupations, was either negative or insignificant because the average
SSI of the “training-based” occupations was lower than (not different from) that of the
“guild-based” occupations in the modern sector.
2.4.4 Examining the Validity of the Empirical Strategy
Does the traditional sector in each census year represent a valid counterfactual that
captures what the religious occupational differentials would have been like in the absence
of state-led modernization? This is violated if state modern projects generate general
equilibrium effects, such as crowding-out of private producers via competition in the
output markets.
36
To examine this possibility, I first observe that the religious
occupational segregation in the traditional sector is similar to what has been documented
in the late eighteenth and early nineteenth century, i.e. before modernization (Raymond
1973, pp. 524-6). Second, the employment share of the modern sector is 8 percent and 3

36
Forced crowding-out through drafting of private producers into state projects is also possible and has
been documented in Owen (2002, p. 69) especially in the case of textile workers.
79

percent of adult active male population in Cairo and Alexandria in 1848 and 1868
respectively. The percentages are even lower in the other provinces at 2 percent and 0.3
percent.
37
These numbers indicate a significant modern sector which is not, however,
large enough to generate general equilibrium effects. This observation is confirmed by
estimating the following panel equation at the district-level:
68ℎ !"

=

+
1868
+ #'"

+

ThreatIndustry is the percentage of active adult males in the traditional sector in the
district who work in a “threatened” industry; i.e. an industry for which a state project has
been established. Projects is the supply of state projects in the “threatened” industry in
the district. is district fixed effects, and is 1868 year fixed effect. I estimate
equation (6) for each “threatened” industry separately, and I conduct the analysis at the
district-level to have a meaningful output market. The results are shown in table 2.6. In
all three industries: textiles, transportation, and military, the coefficient on Projects is
either insignificant or positive, thus implying no crowding-out effects.
But are there significant differences in the religious bias of the recruitment policies
across modern and traditional sectors that may drive the results? I compare the religious
composition within the “overlapping” or “guild-based” occupations across traditional and
modern sectors. My rationale is that if the religious bias of recruitment policies differed
across the two sectors, one should observe different religious composition of the

37
Marsot (1984, p. 181) cites a much higher employment at 20-25 percent of adult males who are at least
15 years old in 1833. If her estimates are accurate, employment must have fallen sharply in the 1840s, and
what one observes in 1848 is already a low point in the employment of modern projects. This might be in
line with the general decline of state-led modernization in response to the London treaty of 1841.
80

“overlapping” occupations, on average. In 1848, the average percentage of Christians in
the “overlapping” occupations in the traditional and modern sectors is 8 percent and 6
percent respectively. In 1868, the percentages are almost identical across the two sectors
(7 percent).
38
In figure 2.3, I show the percentage of Christians in the large (>= 9 workers
in each sector) “overlapping” occupations in both sectors. These occupations constitute
55 percent (53 percent) of modern sector employment in 1848 (1868). Overall, the
religious composition of these occupations is very similar across the two sectors, thus
suggesting no significant difference in the religious bias of recruitment policies.
39

2.4.5 State Industrialization and Religious Occupational Segregation
Were state modern projects more integrated within occupations across religious
groups than the traditional guilds? I measure occupational segregation by two segregation
indexes, dissimilarity and isolation. Dissimilarity index measures the share of the
Christian population that would need to change occupation so that religious groups
become evenly distributed across occupations within the sector (traditional or modern).
Isolation index, on the other hand, measures the percentage of Christians in the
occupation practiced by the average Christian (Cutler and Glaeser 1997).

38
The z-statistic for test of equality of proportions across sectors is 0.005 and -0.003 in 1848 and 1868
respectively. The number of "overlapping" occupations in 1848 and 1868 is 44 and 43 respectively, and is
equal, by construction, across traditional and modern sectors. The computed average percentage of
Christians in each sector is the average of the percentages of Christians over “overlapping” occupations in
that sector, weighted by the occupation size. The occupational percentage of Christians used in calculating
this average is computed from the systematic sample only.
39
The religious bias of recruitment policies in the “training-based” occupations may have been different
from that in the traditional sector. Evidence cited in Heyworth-Dunne (1938, p. 338) suggests that the
government did not hire Copts as printing workers (!). However, the analysis in subsection 2.4.3 shows that
the main findings are driven by technology mechanism which is in turn affected by the recruitment policies
in the “guild-based” occupations only.
81

Table 2.7 shows the occupational segregation indexes for the traditional and modern
sectors in 1848 and 1868, and for each of the “old,” “overlapping,” and “new”
occupations, where “old” occupations are those that exist in the traditional sector only. I
find that occupational segregation is not attenuated by modernization, and that the work
force in state modern projects was equally segregated as the work force in the traditional
sector. Interestingly, occupational segregation is not restricted to “guild-based”
occupations and is equally large in the “training-based” occupations.
The result seems to be consistent with considerable historical evidence that suggests
that recruitment policies of the modern sector generally preserved the (religiously-
segregated) guild structure. The 1823 account books of the largest textile manufactory in
Egypt in 1848 classified workers by guilds (Ghazaleh 1999, p. 131). Ghazaleh (1999, pp.
122-3) cites several events in which the government resorted to the guilds headmen in
order to solve problems with workers in the state projects. Baer (1964, p. 94) states
explicitly that Muhammad Ali used guilds to supply workers for his manufactories.
Along the same lines, an 1876 document in Amin (1928, p. 1304) elaborates the role of
the guilds headmen in resolving conflicts between the state and workers in a modern
industrial institution, Alexandria port. Baer (1964, p. 108) makes a similar point when
mentioning that the steam navigation companies demanded the opening up of the labor
market to get rid of the control of the pilots’ guild over the supply of workers.

82

2.5 Empirical Analysis: Educational and Military Modernization
2.5.1 Educational Modernization
Having examined the impact of state industrialization, I now turn to educational
modernization. I first evaluate the impact of four public modern schools of higher
education: engineering, medicine, veterinarian medicine, and translation that operated
over the period (1827-54), and that did not admit non-Muslims until 1873, on religious
differentials in occupational attainment. I then examine whether modern schools, both
public and private (at all educational levels), that grew since the first half of the
nineteenth century, acted as an integrating device for religious minorities, or they were
rather equally segregated as the traditional educational scheme.
With regard to the first question, I estimate the following linear probability
regression:
7;""

=

+

+ #6!"

+ 8'ℎ

+ , <'ℎ

+
/6!"

× 8'ℎ

+ 36!"

× <'ℎ

+ 4 =

+

In this regression, Professional is an index for working in a professional occupation
(engineer, physician, pharmacist, veterinarian, medical assistant, nurse, translator, or
interpreter). is an 1868 census year dummy. Nonmuslim is an index for being non-
Muslim. Tcohort is an index for the “treated” cohort (1812-34) that was poised to benefit
from the public modern schools operating in 1827-54.
40
Ycohort is an index for the

40
I assume that students enter these schools for a period of five years, from age 15 to 20, which is
consistent with the admission system in these schools (Sami 1928, Vol. 3(3), p. 1123).
83

younger untreated cohort (1834-47) that witnessed the closure of the schools by the time
it was in higher-education age. The base cohort is the older untreated cohort (1778-1812)
that was too old to enter these schools when they were opened. Identifying the effect of
modern schools (and whether it varies by religion) comes from differences in the
availability of the four schools across cohorts of birth. Z is a vector of individual-level
characteristics including slave index, foreigner index, black index, migrant index, and the
interactions of the foreigner index with cohort of birth. I restrict the sample to active
males who are 21-70 years old in each census year, and I pool the two census years
together to obtain a reasonable sample size in each birth cohort in the professional
occupations. Finally, I estimate the regression for the full sample, but when I restrict it to
the urban areas, where the four schools operated, the results are not affected.
Table 2.8, column (1), shows the results. The coefficients should be interpreted as
the lower bound of the true effects of the four schools. Graduates of these schools may
have worked in other professional occupations which are not directly related to the fields
they specialized in.
41
Also, in a society with a very low human capital, providing
education to these pioneering students is expected to have high returns and to create
knowledge spillovers. I find that Muslims in the “treated” cohort are more likely to work
in a professional occupation by 0.4 percentage point than their coreligionists in the old
cohort. Non-Muslims in the old cohort are not statistically different from Muslims
(although the effect is positive at 0.3 percentage point), and the religious “professional”

41
Biographies of prominent graduates of these schools confirm this possibility. Many of these graduates
worked as managers of governmental departments and state production projects, judges, and lawyers
(Heyworth-Dunne, 1938, pp. 159-63, 177-80, 253-64, 269-71, 304-7, 326-9).
84

gap in the treated cohort is smaller, but statistically not different from, the gap in the old
cohort. I find that the young “untreated” cohort is insignificantly different from the old
“untreated” cohort, thus lending support to the hypothesis that it is the four schools,
rather than general trend effects, that explain the upward mobility of Muslims towards
professional occupations in the treated cohort. Historical evidence lends support to this
finding.
42

The finding comes as no surprise given the institutional ban on non-Muslims from
entering these schools. They had to resort to local guilds training, or training outside
Egypt, in order to enter these high-profile occupations. 75 percent of non-Muslims in
professional occupations in the sample are foreigners, compared to 0.6 percent among
Muslims. Thus, it is possible to conclude that the four schools provided upward mobility
routes for local Muslims to replace foreign non-Muslims in professional occupations.
Table 2.9 shows the religious composition of students in religious and modern
schools at both the pre-higher and higher-education levels in 1848 and 1868.
43
First,
modern schools were equally religiously segregated as the religious schools, with
Muslims going to public schools and non-Muslims to private ones. This segregation
pattern persisted until mid-twentieth century. Second, although modern schooling
generally expanded between 1848 and 1868, the growth was faster among non-Muslims.

42
Heyworth-Dunne (1938, p. 433) mentions that “the training of translators has practically ceased with the
closing of the establishment… and the government services had to fall back on Syrians or foreigners.”
43
Sample consists of males currently enrolled in schools in 1848 and 1868 with non-missing religion. Low
enrollment in public schools is consistent with the evidence on the closures of public schools after 1841.
85

40 percent of non-Muslim students went to modern schools in 1868, as opposed to only 4
percent of Muslim students.
2.5.2 Military Modernization
Promoting Egyptians to high commissioned officers’ ranks and conscripting non-
Muslims in the army under Sa’id (1854-63) were two major reforms in the military and
social history of Egypt. In this subsection I examine whether the promotion reform
benefited both Muslims and non-Muslims equally, and whether the army integrated non-
Muslims at all military ranks in 1868, as a result of the reforms, as Khedive Ismail
claimed. In particular, I estimate the following linear probability equation:
8*&;;
'

=

+

+ #6!"

+ 'ℎ

+ /6!"

×
'ℎ

+ 4 =

+

CommOfficer is an index for being a commissioned officer. Mcohort is an index for
the “treated” cohort (1838-47) that was poised to benefit from the promotion reform.
44

The base cohort in the regression is the older cohort that was born in 1778-1838. The
results are shown in table 2.8, column (2). Two caveats in the regression are the missing
age for 32 percent of the commissioned officers in the sample,
45
and the extremely small
number of non-Muslim officers (only 2 officers). Non-Muslims in the old cohort are less
likely to be officers than Muslims. Muslims in the treated cohort are more likely to be

44
Since I could not find a specific date for the first reform, I assume it took place in the middle of Sa’id’s
reign, i.e. in 1858/59. I assume that the promotion to commissioned officer’s rank occurs at 20 years old
which is the average age of graduation from military schools.
45
Military personnel were usually not enumerated in the 1848 and 1868 censuses as they were to be
enumerated in an “undiscovered” military census. Their census returns thus only include name, location,
occupation, and nationality. See Chapter 1 for more details.
86

commissioned officers than their coreligionists in the old cohort, and the gap is even
wider in this cohort in favor of Muslims. Foreigners in the old cohort are more likely to
be officers than locals (since Turks dominated the commissioned officers’ body before
the reform). The foreigner-local gap is not affected by the reform, however.
Because of the small number of non-Muslims in the army and the scarcity of
information on the covariates for the military personnel, I evaluate whether both reforms
led to the integration of non-Muslims in the army by comparing the religious and
nationality composition of the Egyptian army in the sample across 1848 and 1868 in
Table 2.10. The army is overwhelmingly Muslim in both 1848 and 1868, but there is a
rise in the share of non-Muslims in the total military personnel from 1 percent to 3
percent (still below their population share of 7 percent) between the two years, perhaps as
a result of the conscription reform. Although Egyptian non-Muslims had access to
commissioned officers’ ranks by 1868 because of the promotion reform, their
representation in the army remained mainly at the soldier level. Egyptian Muslims, on the
other hand, improved their lot vis-à-vis the Turkish officers from 40 percent of the
officers’ body in 1848 to 69 percent in 1868. Thus, it appears that the officers’ promotion
reform benefited mainly Egyptian Muslims, while the non-Muslims’ conscription reform
increased their access to the army as soldiers in 1868. Contrary to what Khedive Ismail
believed, their representation in the army was still below their national share, thus
suggesting they were not fully integrated yet, especially at the commissioned officers’
ranks.

87

2.6 Conclusion
State-led modernization in nineteenth century Egypt was perhaps the most
significant institutional intervention in the Middle East since the Ottoman Conquest in the
sixteenth century. The state’s intense modernization programs in production, education,
and the army, strongly shook the inherited medieval institutions that preserved deeply-
rooted inequalities, most importantly between religious groups. Using samples that I
digitized of the 1848 and 1868 Egyptian census manuscripts, the earliest individual-level
data source in the Middle East, the chapter sheds entirely new lights onto two major and
highly controversial questions in the Middle Eastern economic history literature. It
provides the first rigorous quantitative evidence on the magnitude of the human capital
gap between Muslims and non-Muslims in Egypt, on the nature of the technologies used
in Egypt’s long-debated industrial modernization program in its two waves, and on the
impact of industrial, educational, and military modernization on the religious gap.
I found that the first wave of industrial modernization in 1816-48 increased the
religious human capital gap. Its technology increased the relative demand for low skills
(primarily in textiles), affected Muslims adversely but benefited Christians who
experienced upward mobility because their administrative skills were highly demanded
by the manufactories. The second wave in 1848-68 did not alter the gap. Its technology
that was complementary to high skills (primarily in transportation), led to upward
mobility among both Muslims and Christians. In contrast, educational and military
modernization benefited Muslims more than non-Muslims, and increased their upward
mobility to high-profile jobs, such as physicians, engineers, translators, and high-ranked
88

commissioned officers in the face of foreign non-Muslims (in professional occupations)
and the Turkish Muslim elite (in military officers’ ranks). Nevertheless, the impact of
these reforms was limited to a tiny fraction of the Egyptian Muslim population and thus
did not lead to a general catching-up effect with non-Muslims. In all three modern
institutions, state industrial establishments, modern schools, and the army, religious
segregation was not attenuated both because the traditional skill-acquisition institutions,
guilds and kuttabs, remained dominant, and because the new channels for mobility (job
training, modern education, military training), that were created by modernization, were
themselves segregated.
The findings of the chapter have implications for several literatures including those
on state-led modernization, technological change and labor market inequality, and ethno-
religious segregation. The failure of the first wave of industrial modernization in Egypt
(1816-48) in raising the human capital of the Muslim population may suggest that a
successful state-led modernization program should establish skill-acquisition institutions
rather than pure production institutions. Focusing on state production projects that still
rely on traditional skill-acquisition institutions makes modernization largely superficial.
My findings also suggest that technological change alters the relative demand for skills
but that this effect may vary across social groups. The effect on each group depends on
the matching between the group’s traditionally-dominated skills and the specific skills
that the modern technology demands. Finally, my findings also imply that ethno-religious
occupational segregation can be altered by a deliberate program if the program provides
89

alternative routes for skill-acquisition through job training, educational, and military
modernization that can effectively replace the traditional gap-preserving institutions.
Convergence in educational and labor market outcomes between religious groups in
Egypt may have had to await the second major (and highly controversial) state-led
modernization program under Nasser (1952-70). I show in chapter 3 that the educational
modernization aspect in Nasser’s program, that made a complete transformation of
kuttabs to public modern schools, was more successful in closing the religious gap by the
end of the twentieth century. In an ongoing research, I am also investigating whether
nineteenth century state industrialization that did not attenuate religious occupational
segregation, reduced religious residential segregation through creating more integrated
working-class neighbourhoods around the state production projects. I am exploring this
possibility using GIS techniques to map the 1848 and 1868 individual census records and
the state projects in Cairo and Alexandria.

90

Figure 2.1: Occupational Distribution by Religious Group in 1848-68
Egypt (Traditional Sector)

0
10
20
30
40
50
60
70
80
90
100
Muslims Christians Jews Muslims Christians
Urban Egypt Rural Egypt
Unskilled Workers
Skilled Workers
White-Collar Workers

Sample is restricted to males employed in the traditional production sector who are at least 15 years old,
including those with missing age. Statistics are based on the systematic sample only, and are weighted to
account for the different sampling rates across the provinces. Each column represents the share of the
occupational category within the relevant religious group. Urban Egypt is defined as Cairo and Alexandria,
while rural Egypt includes all the other provinces.

91

Figure 2.2: State Industrialization and Christian-Muslim SSI Gap
(Christian-Muslim Differences in Ordered Logit Predicted Probabilities for Each SSI
Value)

-0.4
-0.3
-0.2
-0.1
0
0.1
0.2
0.3
0.4
1 2 3 4 5 6 7 8 9 10 11 12
Difference in Probability
SSI
A. 1848
Traditional Sector
Modern Sector

-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
1 2 3 4 5 6 7 8 9 10 11 12
Difference in Probability
SSI
B. 1868
Traditional Sector
Modern Sector

Christian-Muslim differences in predicted probabilities of SSI values in each sector are based on the
estimates of the ordered logit regressions of table 2.3, columns (1) and (3). Predicted probabilities are
calculated for a free, non-black, non-foreigner, and non-migrant individual, while all the other regressors
are set at their mean values. See Appendix C for definitions of SSI values. The predicted probabilities of
the 12 SSI values for a Muslim in the traditional sector in 1848 are 0.01, 0.1, 0.02, 0.13, 0.36, 0.20, 0.05,
0.04, 0.07, 0.01, 0.01, and 0.00. In 1868, they are 0.00, 0.11, 0.02, 0.22, 0.12, 0.27, 0.08, 0.05, 0.11, 0.01,
0.01, and 0.01. All differences in predicted probabilities are statistically significant at the 5 percent level in
1848. In 1868, they are all significant except at SSI=6 in the traditional sector and at SSI=7 in the modern
sector. All the difference-in-differences estimates are statistically significant at the 5 percent level in both
years, where the standard errors are estimated using the delta method.
92

Figure 2.3: Percentages of Christians in the Large
"Overlapping" Occupations in Traditional and Modern Sectors
0
0.1
0.2
0.3
0.4
0.5
0.6
A. 1848
Traditional
Modern

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
B. 1868
Traditional
Modern

Sample is restricted to males who are employed in the “overlapping” occupations in the traditional and
modern sectors, and are at least 15 years old, including those with missing age. Statistics are based on the
systematic sample only. A large occupation is an occupation with at least 9 workers in each of the
traditional and modern sectors. Occupations are ordered in the figure by the occupation’s size in the
modern sector, from the largest to the smallest.
93

Table 2.1: Descriptive Statistics: Modernization, Religion, and
Human Capital in 1848 and 1868

1848 1868
Cairo and
Alexandria
Other
Provinces
Cairo and
Alexandria
Other
Provinces
Employment Status N=11,324 N=12,457 N=19,110 N=5,878
Employed (%) 86.11 72.57 80.20 68.36
Unemployed (%) 3.26 0.37 4.51 0.41
Out of labor force (%) 2.83 0.04 3.00 0.43
Missing (%) 7.80 27.02 12.29 30.76
Occupational Attainment N=9,792 N=9,040 N=15,286 N=4,021
White-collar worker (%) 21.96 7.73 23.69 9.80
Skilled worker (%) 27.80 6.73 29.55 8.26
Unskilled worker (%) 50.24 85.54 46.76 81.94
Social status index (Mean) 5.43 5.01 5.67 5.21
Modernization N=9,792 N=9,040 N=15,286 N=5,878
Employed in the modern sector (%) 7.55 1.75 3.33 0.32
Military personnel (%) 4.15 1.50 2.31 6.53
Professional occupation (%) 0.90 0.00 0.85 0.00
Religion N=11,324 N=12,457 N=19,110 N=5,878
Muslim (%) 89.90 92.65 86.44 89.79
Christian (%) 7.14 6.34 6.06 6.74
Jew (%) 1.16 0.03 1.01 0.00
Non-Muslim (unspecified) (%) 0.31 0.02 3.43 0.00
Missing (%) 1.07 0.96 3.07 3.47
Demographic
Age (Mean) 37.46
(N=10,593)
40.45
(N=12,294)
35.51
(N=17,071)
38.18
(N=5,858)
Slave or emancipated slave (%) 1.56
(N=11,316)
1.33
(N=12,454)
1.42
(N=19,024)
1.97
(N=5,742)
Outside government's control (%) 10.99
(N=11,018)
1.13
(N=12,419)
12.26
(N=19,004)
0.68
(N=5,864)
Abyssinian, black, or Nubian (%) 3.35
(N=10,794)
0.12
(N=12,336)
5.56
(N=17,852)
2.94
(N=5,846)
Migrant (%) 58.36
(N=9,948)
6.23
(N=12,286)
35.26
(N=18,369)
3.67
(N=5,644)
Sample is restricted to adult males who are at least 15 years old, including those with missing age.
Sample size varies from one variable to another to reflect the missing values for each variable.
Sample weights are used to adjust for the different sampling rates across provinces. Descriptives
are based on systematic sample only. See Appendix C for definitions of variables.

94

Table 2.2: Testing for Technology-Skill Complementarity in State
Production Projects
(Ordered Logit Regression- Dependent Variable: Social Status Index)

1848 1868
All Egypt Cairo and
Alexandria
Other
Provinces
All Egypt Cairo and
Alexandria
Panel A: Overall Technology-Skill Complementarity
Modern 0.254***
(-5.28)
0.356***
(-4.60)
0.083***
(-7.04)
2.021***
(3.94)
1.994***
(4.26)
District fixed effects Yes Yes Yes Yes Yes
Pseudo R-squared 0.044 0.043 0.040 0.029 0.03
Panel B: Technology-Skill Complementarity by Industry
Textiles × Modern 0.018***
(-15.14)
0.039***
(-9.02)
0.006***
(-12.63)
0.072**
(-2.53)
0.089***
(-2.61)
Transportation × Modern 2.063
(0.49)
2.537
(0.45)
3.064
(0.63)
24.948***
(14.45)
16.865***
(12.63)
Military × Modern 0.950
(-0.20)
1.154
(0.60)
0.722
(-0.32)
0.590
(-1.35)
0.974
(-0.13)
Other Modernized Industries
× Modern
0.265**
(-2.46)
0.401**
(-2.15)
0.028***
(-2.86)
0.755
(-0.65)
0.749
(-0.71)
Textiles 2.041***
(5.51)
0.922
(-0.46)
7.562***
(24.65)
1.403***
(3.94)
1.234**
(2.11)
Transportation 0.081***
(-14.11)
0.168***
(0.033)
0.006***
(-16.35)
0.132***
(-16.81)
0.183***
(-14.32)
Military 1.663***
(3.96)
0.920
(-0.73)
7.284***
(13.23)
1.127
(1.34)
1.003
(0.04)
Other Modernized Industries 0.761***
(-2.73)
0.609***
(-5.33)
2.622***
(2.99)
0.738***
(-4.08)
0.679***
(-5.28)
District fixed effects Yes Yes Yes Yes Yes
Pseudo R-squared 0.073 0.067 0.093 0.051 0.051
# Districts 97 15 82 53 19
# Villages/Urban quarters 859 194 665 616 250
# Observations 18,146 9,315 8,831 18,968 15,133
Odds ratios are reported. Z-values are between parentheses. Robust standard errors are clustered at
the village/urban quarter level. Sample is restricted to employed males who are at least 15 years
old with non-missing values for all variables included in the regression. I do not run the regression
for the other provinces in 1868 because the number of workers in the modern sector in the sample
is very small (13 workers). Controls in each regression are: age, foreigner dummy, slave dummy,
migrant dummy, and black dummy. * indicates significance at 10 percent level, ** indicates
significance at 5 percent level, and *** indicates significance at 1 percent level.

95

Table 2.3: Impact of State Industrialization on Inter-religious
Occupational Differentials- Ordered Logit Model
(Dependent Variable: Social Status Index- Factor Changes in Odds Ratios are
Reported)
1848 1868
Christian 2.833***
(8.28)
2.992***
(7.79)
3.137***
(10.48)
2.889***
(10.38)
Jew 1.268
(0.76)
1.303
(0.86)
1.594***
(2.62)
1.459**
(2.10)
Modern 0.183***
(-6.93)
0.987
(-0.05)
1.742***
(2.67)
0.477***
(-2.52)
Christian × Modern 12.703***
(8.60)
2.393**
(2.55)
1.602
(1.42)
6.248***
(3.43)
Textiles × Modern 0.020***
(-10.29)

Textiles × Modern × Christian 51.432***
(6.67)

Textiles 2.575***
(9.23)

Textiles × Christian 0.388***
(-3.87)

Transportation × Modern 45.476***
(10.03)
Transportation × Modern × Christian 0.190**
(-2.23)
Transportation 0.141***
(-16.85)
Transportation × Christian 0.740
(-0.78)
District fixed effects
Yes Yes Yes Yes
Pseudo R-squared
0.053 0.059 0.033 0.053
# Observations
18,039 18,039 18,526 18,526
Factor changes in odds ratios are reported. Robust Z-values, clustered at the village/urban
quarter level, are in parentheses. Sample is restricted to active males who are at least 15 years
old with non-missing values for all variables included in the regression. Additional controls
in each regression are: age, foreigner dummy, slave dummy, migrant dummy, and black
dummy. * indicates significance at 10 percent level, ** indicates significance at 5 percent
level, and *** indicates significance at 1 percent level.

96

Table 2.4: Impact of State Industrialization on Inter-
religious Occupational Differentials- Multinomial
Logit Model
(Dependent Variable: Occupational Group- Partial Derivatives are
Reported)

(1) Outcome:
Unskilled
Worker
(2)
Outcome:
Skilled
Worker
(3)
Outcome:
White-
Collar
Worker
Panel A: 1848
Christian -0.387***
(0.031)
0.107***
(0.023)
0.276***
(0.032)
Jew -0.111**
(0.058)
0.023
(0.014)
0.088
(0.058)
Modern 0.052
(0.021)
0.002
(0.014)
-0.054***
(0.012)
Christian × Modern -0.361
(0.131)
-0.012
(0.028)
0.371***
(0.115)
District fixed effects
Yes Yes Yes
Predicted Probability of Outcome
0.806 0.076 0.113
Pseudo R-squared
0.190 0.190 0.190
# Observations
18,039 18,039 18,039
Panel B: 1868
Christian -0.436***
(0.026)
0.141***
(0.025)
0.295***
(0.033)
Jew -0.175***
(0.047)
0.014
(0.062)
0.161**
(0.067)
Modern -0.214***
(0.039)
-0.007
(0.027)
0.222***
(0.035)
Christian × Modern -0.092
(0.154)
-0.024
(0.080)
0.115
(0.136)
District fixed effects
Yes Yes Yes
Predicted Probability of Outcome
0.600 0.202 0.198
Pseudo R-squared
0.130 0.130 0.130
# Observations
18,526 18,526 18,526

All columns in each year are from the same multinomial logit regression. Partial
derivatives are reported. Standard errors, calculated by the delta method, are in
parentheses, where the original robust standard errors are clustered at the
village/urban quarter level. Sample is restricted to employed males who are at
least 15 years old with non-missing values for all variables included in the
regression. Additional controls in each regression are: age, foreigner dummy,
slave dummy, migrant dummy, and black dummy. * indicates significance at 10
percent level, ** indicates significance at 5 percent level, and *** indicates
significance at 1 percent level.

97

Table 2.5: Mechanisms of State
Industrialization: Technology or Job
Training?
(Ordered Logit Regression- Dependent Variable:
Social Status Index)
1848 1868
Christian 2.855***
(8.20)
3.134***
(10.44)
Jew 1.241
(0.68)
1.589**
(2.60)
Modern 0.429***
(-3.09)
1.913
(2.92)
Modern × New Occupation 0.070***
(-8.39)
0.757
(-0.70)
Christian × Modern 6.629***
(6.27)
2.816***
(3.46)
Christian × Modern × New Occupation 0.055***
(-7.54)
0.180***
(-3.29)
District fixed effects
Yes Yes
Pseudo R-squared
0.056 0.033
# Observations
18,039 18,526

Odds ratios are reported. Z-values are between parentheses.
Robust standard errors are clustered at the village/urban quarter
level. Sample is restricted to employed males who are at least 15
years old with non-missing values for all variables included in
the regression. Additional controls in each regression are: age,
foreigner dummy, slave dummy, migrant dummy, and black
dummy. * indicates significance at 10 percent level, ** indicates
significance at 5 percent level, and *** indicates significance at
1 percent level.

98

Table 2.6: Crowding-out Effects of State
Industrialization
(Fixed Effects Panel Regression- Dependent Variable:
Percentage in a Threatened Industry)

Textiles Transportation Military
Projects -0.004
(0.010)
-0.008
(0.042)
0.028**
(0.012)
1868 Effect -0.004
(0.005)
0.021
(0.023)
-0.002
(0.003)
Constant 0.024
(0.003)
0.044
(0.012)
0.004
(0.002)
R-squared (Overall) 0.003 0.005 0.256
# Districts 150 150 150

Standard errors are in parentheses. Percentage in a threatened
industry is calculated out of the employed adult male population who
are at least 15 years old in the traditional sector in the district.
Systematic sample only is used in the calculation. * indicates
significance at 10 percent level, ** indicates significance at 5
percent level, and *** indicates significance at 1 percent level.

Table 2.7: Occupational Segregation in Traditional and Modern
Production Sectors
Traditional Modern
Old Overlapping Overall Overlapping New Overall
1848
Dissimilarity 0.423 0.756 0.525 0.749 0.06 0.635
Isolation 0.932 0.919 0.928 0.948 0.969 0.953
# Occupations 180 44 224 44 4 48
1868
Dissimilarity 0.433 0.664 0.528 0.82 0.721 0.786
Isolation 0.946 0.93 0.939 0.928 0.919 0.925
# Occupations 162 43 205 43 15 58
"Old" occupations are the occupations that exist only in the traditional sector. "Overlapping" and
"New" occupations are defined as in the text. Only systematic sample of employed adult males who
are at least 15 years old with non-missing religion is used in the calculations, and dataset is
collapsed at the occupation-level.

99

Table 2.8: Educational and Military Reforms
and Religious Occupational Differentials
(Linear Probability Regression)

(1)
Dependent
Variable =
Professional
Index
(2)
Dependent
Variable =
Commissioned
Officer Index
Non-Muslim 0.003
(0.002)
-0.009***
(0.002)
Born in 1812-34 0.004***
(0.001)

Born in 1834-47 -0.000
(0.001)

Born in 1838-47 0.005**
(0.002)
Non-Muslim × Born in 1812-34 -0.006
(0.004)

Non-Muslim × Born in 1834-47 0.001
(0.006)

Non-Muslim × Born in 1838-47 -0.008*
(0.004)
Black -0.004
(0.004)
-0.006***
(0.002)
Slave 0.004
(0.010)
-0.004**
(0.002)
Foreigner 0.009
(0.011)
0.016***
(0.005)
Foreigner × Born in 1812-34 -0.014
(0.009)

Foreigner × Born in 1834-47 -0.014
(0.011)

Foreigner × Born in 1838-47 -0.000
(0.010)
Migrant 0.005**
(0.003)
0.004**
(0.002)
1868 Index 0.003
(0.002)
0.005***
(0.001)
Constant 0.002
(0.001)
-0.000
(0.001)
Adjusted R-squared 0.012 0.01
District fixed effects Yes Yes
# Observations 29,572 29,572
Robust standard errors that are clustered at the village/urban quarter level are between parentheses. Sample
is restricted to employed males who are 21-70 years old in each census year with non-missing values for all
the variables included in the regressions. The two census samples are pooled together. * indicates
significance at 10 percent level, ** indicates significance at 5 percent level, and *** indicates significance
at 1 percent level.
100

Table 2.9: Religious Composition of Religious and Modern
Schools in 1848 and 1868
(M: Muslim Students; NM: Non-Muslim Students)

Religious Schools Modern Schools

Muslim
Schools
Non-Muslim
Schools
Public Schools Private
Schools
M NM M NM M NM M NM
1848
Pre-Higher
Education 692 0 0 280 6 0 0 0
Higher Education 111 0 0 0 7 0 0 0
Total 803 0 0 280 13 0 0 0
1868
Pre-Higher
Education 1,561 0 0 115 42 0 6 76
Higher Education 266 0 0 1 23 0 0 0
Total 1,827 0 0 116 65 0 6 76
Sample is restricted to male students who are currently enrolled in schools in the 1848 and
1868 censuses with non-missing religion. School sector and level for modern schools are
inferred from school names which are usually recorded in the census records, combined with
information on schools from Heyworth-Dunne (1938).

101

Table 2.10: The Composition of the Egyptian
Regular Army by Religion and Nationality in 1848
and 1868
(E: Egyptians; F: Foreigners; U: Unknown Nationality; T: Total)

Muslim Military
Personnel
Non-Muslim
Military Personnel

E F U T E F U T
1848
Soldier 310 22 68 400 3 0 0 3
Non-Commissioned Officer 23 10 3 36 1 0 0 1
Commissioned Officer 24 16 20 60 0 0 0 0
Total 357 48 91 496 4 0 0 4
1868
Soldier 396 8 0 404 13 1 0 14
Non-Commissioned Officer 11 1 1 13 0 0 0 0
Commissioned Officer 97 40 4 141 2 0 0 2
Total 504 49 5 558 15 1 0 16
Sample is restricted to military personnel who are 21-70 years old with non-
missing religion, but including those with missing age. There are 27 (7)
military personnel in the sample with missing religion in 1848 (1868). Their
occupational distribution in 1848 (1868) is as follows: 20 (3) soldiers, 1 (0)
non-commissioned officer, and 6 (4) commissioned officers.
102

Chapter 3

From Kuttabs to Schools: Educational Modernization,
Religion, and Human Capital in Twentieth Century Egypt

3.1 Introduction
When the English took Egyptian affairs in hand, the accountants in the
employment of the Egyptian government were almost exclusively Copts.

Lord Cromer, British Consul-General of Egypt, 1908
1

An intriguing phenomenon in the economic history of the Middle East is that non-
Muslim minorities enjoyed on average better educational and occupational outcomes than
the Muslim majority over the course of the twentieth century. The above statement by
Lord Cromer, made upon the British occupation of Egypt in 1882, and hinting at the
different occupational outcomes of the largest non-Muslim minority in Egypt, the Copts
(Egyptian Christians), is but one example of the large anecdotal evidence in support of
the existence of this phenomenon. The phenomenon, which is observed with some
variations across many countries in the Middle East, is attested to in various historical
sources (Tagher 1998, Courbage and Fargues 1997, Issawi 1981, Krämer 1989), and also
seems to have quantitative support from the available data sources. In 1986 Egypt, one of
the largest Middle Eastern countries, the average literacy rate for Muslim males born in

1
Tagher (1998), p. 212.
103

(1910-1920) was 32.2 percent, while average literacy for Christian males born in the
same cohort was much higher at 49.4 percent. For the younger cohort born in (1970-76),
the average literacy for Muslims was 85.6 percent, while that of Christians was 89
percent, suggesting a smaller yet still existent literacy gap (Figure 3.1). With respect to
occupational outcomes, in 1986, about 3 percent of male Muslims and 10.1 percent of
male Christians born in (1910-1920) were working in white- collar occupations. For the
cohort born in (1950-1959), the percentages were 29.5 percent and 47.5 percent for
Muslims and Christians respectively, suggesting a persistent (or even wider) occupational
gap (Figure 3.2).
2

The expansion in the supply of modern schools is one of the main forces that can
cause abrupt changes in the educational and occupational distributions of a given
population (Duflo 2001; Yuchtman 2010), and can hence alter the inter-religious human
capital differentials. Over the course of the twentieth century, modern schooling in Egypt
has largely expanded to replace the traditional religious schools or kuttabs that were the
largest provider of education in the country up to mid- twentieth century. In 1907, while
an Egyptian district had 0.25 modern schools (per 10,000 individuals), on average, the
average number of kuttabs (per 10,000 individuals) was about 3.6. By 1959, the picture
was reversed as all the traditional kuttabs were officially transformed into modern

2
Author’s calculations based on the individual- level 10 percent sample from the 1986 census. I will
describe the data sources in section 3.3.
104

schools. In 1959, the average number of modern schools per 10,000 individuals in a
given Egyptian district has increased to 3.6, while kuttabs have totally vanished.
3

I examine the impact of the expansion of modern schooling on the educational and
occupational differences between Muslims, Christians, and Jews over the twentieth
century, using several new data sources from Egypt. In particular, I attempt to answer the
question: Did the expansion of modern schooling (measured by the local supply of
schools at the district- or province- level) lead to a reduction or rather an aggravation of
the educational and occupational differences across Muslims, Christians, and Jews? I
attempt to find out the mechanisms that led to the effectiveness of a specific expansion
program and the less effectiveness of another program. The analysis draws on the
following data sources: (i) village/urban quarter- level census data for the period (1897-
1986), (ii) individual-level 10 percent samples from the 1986 census, and (iii) the regular
schooling reports issued by the Egyptian government over the period (1907-1969).
There is a growing body of economics literature on the relationship between religion
and economic outcomes (Barro and McCleary 2003 and 2005, Guiso et al. 2003, La Porta
et al. 1997, Botticinni and Eckstein 2002 and 2005, Becker and Woessmann 2009,
Boppart et al. 2008, Boorooah and Iyer 2005, Chaudhary and Rubin 2009). This chapter
contributes to this body of literature by examining the aforementioned relationship in the
Middle East, a region where religious identity is the main source of segmentation and
division in the society. In the long run, these religious divisions can have detrimental

3
Author’s calculations based on the regular official schooling reports of 1907/08 and 1959/60. I will
describe this data source in section 3.3.
105

effects as they can lead to tensions that might reach civil war (e.g. Lebanon 1975-1990).
Hence, the Middle East seems to provide a suitable environment to examine the inter-
religious differentials in socioeconomic outcomes.
The major difference between the current chapter and the aforementioned body of
literature, however, is that the previous literature interpreted the inter-religious
differentials in economic outcomes as an impact of religion per se. For example, it has
been argued that specific features of a religion (such as the push for literacy in
Protestantism and Rabbi Judaism) led to higher human capital of the followers of these
religions. By contrast, I do not argue that the observed correlation between religion and
economic outcomes in twentieth century Egypt is due to factors related to religion per se
for two reasons: First, while one may give such interpretation for the Jewish case based
on the literacy requirement under Judaism, it is much harder to make a similar argument
for the Coptic case (who are the vast majority of non-Muslims in Egypt), since there was
nothing of a literacy requirement in the Coptic Church (In fact illiteracy among Copts
was as high in 1986 as 34.62 percent). Second, in chapter 4, I introduce an alternative
economic explanation for the Coptic-Muslim human capital differentials in Egypt by
examining their medieval origins. In particular, I argue that the poll- tax imposed on non-
Muslims upon the Islamic conquest of Egypt in 641 AD induced the poorer and
uneducated Copts (who constituted the vast majority of the Egyptian population then) to
convert to Islam to avoid paying the tax. This selected conversion led to the shrinking of
the remaining Copts into a Bourgeoisie that was on average better educated and relatively
106

more concentrated in higher socioeconomic occupations than the growing Muslim
majority.
The chapter also contributes to the literature on the impact of the expansion in the
supply of schools on improving economic outcomes (Duflo 2001) and on the shift from
traditional schools to modern schools (Yuchtman 2010). To the best of my knowledge,
this is the first research that extends this literature to examine the impact of the expansion
of the local supply of schools on the inter-religious educational and occupational
differentials. The length of the time period studied in the chapter (1897-1996) allows
evaluating different waves of expansion programs in schooling with different scope and
characteristics. Hence, the results have important implications for the role of educational
policies on reducing (or aggravating) the religious differentials in the society. They may
also shed light on the underlying mechanisms behind the persistence of religious
differentials with respect to specific outcomes as of the end of the twentieth century and
the failure of public policy in reducing these gaps. In this context, Egypt provides an
interesting case study: It is a country which started its attempts at modernization in the
early nineteenth century and took important steps towards state building and elimination
of religious gaps since the second half of the nineteenth century, yet it still experiences
religious tensions (partly due to religious socioeconomic differentials) in the first decade
of the twenty-first century.
The rest of this chapter is organized as follows: Section 3.2 provides an overview of
the history of education in Egypt over the twentieth century and the main schooling
107

expansion waves that took place. Section 3.3 discusses the data. Section 3.4 discusses the
empirical strategy and the results. Finally, section 3.5 concludes.
3.2 History of Education in Twentieth Century Egypt: An Overview
In the early nineteenth century, kuttabs were the sole provider of schooling in Egypt.
These schools were religious in nature, and hence were mostly segregated by religion,
and funded by the religious endowments (Waqf system). Traditionally, the curriculum in
Muslim kuttabs was focused on “learning the orthography of the Arabic language through
memorizing the Quran” (Heyworth-Dunne 1938). Arithmetic was not generally taught in
these schools. Nevertheless, at the end of the course of study, which typically took two to
three years, the student was sent to the Qabbani (market weigher) to learn the basics of
measures and weights. By contrast, historically, students in Coptic schools learned
“religion, good manners, and to read and write Arabic and Coptic.” They also memorized
the Psalms and St. Paul’s Epistles. Arithmetic and geometry were taught in the Coptic
schools to be able to measure the Nile level and land areas, perhaps in order to prepare
the children for their future careers as land surveyors, scribes, and tax collectors.
4
Finally,
Jewish kuttabs were designed to prepare the children for their future occupations as
bankers, moneychangers, and merchants. According to Heyworth-Dunne (1938), Hebrew
in these schools was not only taught for religious purposes.

4
Heyworth-Dunne (1938) mentions the following passage as the earliest account on Coptic schools written
by Sadlier (1693): “... the children were taught religion, good manners, to read and write Arabic and
Coptic... and were taught geometry and arithmetic because these two sciences are very useful and
necessary on account of the overflowing of the Nile, whereby the limits are lost; so that it becomes
necessary for them to measure out their land, and by the benefit of the first of these sciences they compute
the yearly increase.” (Heyworth-Dunne 1938, pp. 85; italics are mine to identify the words of Sadlier
1693).
108

Modern European-style schools gradually increased over the nineteenth and the first
half of the twentieth centuries. There were at least three types of modern schools:
5
(i)
Public modern schools were initiated by Muhammad Ali (1805-1848) in order to meet
the human capital needs of the army, but slowed down after London Treaty 1841 which
imposed limitations on the Egyptian army size. They continued to increase modestly over
the nineteenth and early twentieth centuries. In 1907, however, fees were imposed in
public schools at all levels of education, which may have limited enrollment at these
schools. (ii) Private foreign schools, which were either secular or missionary, emerged at
the hands of the European missionaries in the late eighteenth century and accelerated
since the 1840s, and especially after the British occupation (1882). (iii) Private native
schools, which were initiated by the native non-Muslim communities (Copts and
Egyptian Jews) to replace the traditional kuttabs around mid- nineteenth century and
continued to increase in number over the second half of the nineteenth century and the
first half of the twentieth century to satisfy the needs of these communities. The growth
of these schools along with private foreign schools seemed to have absorbed the vast
majority of native non-Muslims communities and therefore their dependency on
traditional kuttabs declined.
For the Muslim majority, however, the dichotomy of the educational system into
kuttabs and modern schools was the main challenge in the evolution of schooling in
Egypt over the nineteenth and twentieth centuries. While modern schools led the student

5
This section is based on Heyworth-Dunne (1938), Harby (1960), Cochran (1986), Boktor (1936 and
1963), and Girgis (1963).
109

into higher levels of education and hence eventually into prestigious governmental or
non-governmental jobs, kuttabs were not allowing the student to be promoted into
modern higher educational levels (except promotion into higher religious institutions
such as Al-Azhar). In particular, at finishing the kuttab’s course, the student was usually
too old to join the modern primary school and hence had to end up working in the fields.
Although the government realized this challenge as early as in the 1860s, the full
unification of the two educational systems occurred only in 1953. There were various
attempts, however, over the second half of the nineteenth century and the first half of the
twentieth century to modernize the kuttabs. Such attempts included placing the kuttabs
under central government supervision (instead of private Waqfs), which started in 1869,
or under provincial councils (1909-1953). In 1916, the government started to transform
kuttabs into four-grade elementary schools with a more modernized curriculum. These
elementary schools were a “modernized” version of the kuttabs but were still separate
from the modern primary schools and hence were not promoting the student into higher
levels of education. In 1944, tuition fees on public primary schools were abolished, and
in 1949, the curricula of the elementary schools and public primary schools were unified
except for the foreign language requirement in the third grade of primary schools. Despite
all these efforts, the dichotomy of the two types continued. It was only in 1953, that all
the elementary schools were transformed into public primary schools, and the dichotomy
ended.
Figure 3.3 shows the evolution of the average number of kuttabs (elementary
schools) and modern schools (defined as non-elementary schools) per 10,000 individuals
110

across districts over the period (1907-1969). The graph is based on the official schooling
reports in the given years. Schools and kuttabs supply is adjusted by population figures
from the closest decennial population census. The evolution of the supply of modern
schools suggests that there were three major waves of expansion: (i) 1907-1921: Average
number of modern schools increased from 0.25 to 2.43 schools per 10,000 individuals
(more than 8- fold increase). (ii) 1927-1936: Average number of modern schools
increased from 1.83 to 2.68 schools per 10,000 individuals (by 46.45 percent). (iii) 1948-
1959: Average number of modern schools increased from 2.11 to 3.6 schools per 10,000
individuals (by 70.62 percent). Rapid population growth between 1948 and 1960 by
about 38 percent has to be put into account when judging the relative magnitude of the
1950s expansion. Supply of kuttabs (elementary schools), on the other hand, shows
modest expansion over the first half of the twentieth century despite the deliberate
expansion of elementary schools that took place in the 1920s to eradicate illiteracy.
6

Two notes are in order: First, while modern schools in the first and second waves
were non-free (whether public or private, foreign or native), they were free in the 1950s
expansion of schooling. As mentioned previously, the fees on public primary schools
were abolished in 1942, but the relative limited supply of these schools in the 1940s may
have weakened the impact of that policy. Second, the gradual expansion of elementary
schools over the twentieth century was directed at eliminating illiteracy. However, the
low expected returns to enrollment in these schools, which did not allow the student to

6
This expansion was an application of the 1923 constitution that provided for free and compulsory
elementary education for every child between 6 and 12 years.
111

transfer to higher levels of schooling, may have reduced the demand for enrollment in
these schools. On the contrary, modern schools were perhaps a more attractive option for
the children since they offered a higher expected return to education that could offset the
cost of education (tuition fees and opportunity cost). Therefore, one can conclude that,
holding the cost of education constant, the expansion of elementary schools may have
been less decisive in eliminating illiteracy than the expansion of modern schools. In
particular, the 1950s expansion with its free modern schools may have been more
effective in combating illiteracy than the earlier expansion of free elementary schools.
3.3 Data
I employ three new data sources to examine the impact of the expansion of modern
schooling on the human capital differences between religious groups. These sources are:
1. Village/Urban Quarter- Level Census Data (1897-1986): These data are at the
lowest administrative level in Egypt, the rural village or urban quarter, which is roughly
equivalent to a zip code in the U.S. The data come from the published census reports of
the decennial censuses: 1897, 1907, 1917, 1927, 1937, 1947, 1960, 1976, and 1986. They
were digitized by Centre d’Etudes et de Documentation Economiques, Juridiques, et
Sociales (CEDEJ) in Cairo, Egypt and made publicly available on a CD-ROM.
The specific data collected differed from one census year to another on an ad hoc
basis. I checked all the censuses for internal consistency before employing them in my
analysis and minor data cleaning was involved. Table 3.1 shows the summary statistics
for the key variables of interest by year (across all villages/urban quarters in a given
year). The average percentage of non-Muslims went down slightly from 7 percent in
112

1897 to 5 percent in 1986. The vast majority of non-Muslims are Christians of whom the
Copts (Egyptian Christians) constitute the largest majority. The average percentage of
Jews went down substantially starting from 1960 census (in the aftermath of the creation
of Israel and the Arab-Israeli conflict) as most Jews emigrated from Egypt. Average
percentage of males who are able to read and write increased gradually over the century
but did not reach universal literacy. There is also a gradual rise in the average percentage
of males who work in non-agricultural activities.
Table 3.2 shows the summary statistics by region, excluding the scarcely populated
border provinces, across the two years 1897 and 1986. The average percentage of non-
Muslims (across villages/urban quarters) is the highest in Middle Upper Egypt, where
they are mostly Coptic Christians. Urban Provinces have the second highest share,
followed by Southern Upper Egypt, Northern Upper Egypt, and finally Lower Egypt.
There is a slight decline in the shares of non-Muslims in almost all regions over the
century, but the decline is much sharper in Urban Provinces perhaps due to the massive
emigration of foreigners and non-Coptic Christians in the 1950s and 1960s. As for male
literacy, one can see that Urban Provinces have the highest average male literacy rate,
followed by Lower Egypt, Northern Upper Egypt, Middle Upper Egypt, and then
Southern Upper Egypt. A similar regional pattern exists for the occupational distribution
measures, with the exception of Southern Upper Egypt, which seems to have a relatively
high percentage of males working in non-agricultural activities, or in white-collar jobs,
on average, perhaps because of the thriving tourism industry in this region.
113

I employ these village/quarter- level census data as a pooled dataset as well as a
panel dataset. For the purpose of building a panel, which follows the same geographical
unit (village or urban quarter) over the twentieth century, I started by using the matching
to the 1996 census geographical codes that was already available in CEDEJ dataset. This
matching assigned to every unit the geographical code from the 1996 census if the unit
exists under the same name in 1996. For the units that do not have a match in 1996
census, I made an extensive manual revision by attempting to match every unit of these
across the years in which it appears with exactly the same name (in the same or
neighboring province) following a general rule: I assigned for the geographical unit the
code from the latest census in which it appeared as a “single” geographical unit. Three
situations arose: First, if a geographical unit has been divided at some point in the
century, I collapsed the emerging subsequent sub-units into this older single unit and
assigned to them the code of that unit in the older year. This situation appeared mainly
for the provincial capital cities, which were always treated as one “village/urban quarter”
in the earlier censuses, but were then “promoted” into “districts” with subdividing them
into urban quarters. It also appeared for the Suez Canal cities, which were also treated as
single units in the early twentieth century but were then promoted into districts, and later
into provinces. Second, if one or more geographical units have been united into one at
some point, I collapsed the earlier units into this subsequent later single unit. This
situation was rarer in practice. Third, if a unit disappeared or appeared at some point,
without being part of a unified set of units or a divided preceding unit, I assigned to it the
geographical code of the latest census in which it appears.
114

Although the matching procedure might seem complicated, in practice, it was
straightforward. About 97 percent of the villages/urban quarters in all years had an exact
mapping in 1996 census. The quality of the 1996 census mapping increases in later
censuses (from 94 percent in 1897 to 99 percent in 1986). Also, the quality of matching
in rural provinces is much higher than in urban ones (98 percent as compared to 87
percent). The quality of matching was particularly low (about 67 percent) in urban
provinces in early years (1897 and 1917). The manual revision I made, hence, focused on
the unmatched units, which were about 3 percent of the total number of units.
Nevertheless, modifications to the 1996 mapping to correct for the aforementioned three
situations were necessary in order to obtain consistent mapping for all units. All these
manual modifications are documented in the dataset.
2. The 1986 Individual- Level 10 percent Census Sample: This data source is
available from IPUMS- International. I focus on Muslim and Christian males who were
born in or after 1910. Table 3.3 shows the descriptive statistics.
3. National Official Schooling Reports (1907-1969): I collected information on the
local supply of schools from the regular official schooling reports. These reports were
irregular in the period (1906-1914), and then became triennial in the period (1921-1952).
After the 1952 military coup, they were re-instated starting from 1955, and seem to have
become annual since then. There is considerable variation in terms of the level of data
across the years: In the earliest reports (1906-1914), the reports give detailed data at the
school level for modern schools, but information on kuttabs is very approximate and is at
the province- level. In the reports (1921-1952), the reports give information on the
115

number of elementary schools and modern (non-elementary) schools at the district- level
(which is higher than the village/quarter- level but lower than the province- level). They
also give detailed information on the supply of each type of elementary and modern
schools at the district- or province- level. Finally, the post- 1952 reports give information
on the number of modern schools (broken down by level of education)at the province-
level. I focus on the reports for the years that are close to the census years: 1907, 1921,
1927, 1936, 1948, 1959, and 1969. Figure 3.3 shows the basic trends of the supply of
modern and elementary schools.
3.4 Empirical Strategy and Results
The empirical analysis is divided into two parts: First, I analyze the evolution of the
inter-religious differentials in educational and occupational outcomes using both the
village/quarter- level data (1897-1986) and the individual- level 1986 data. This analysis
will allow me to identify which gaps were closing over time and which ones were more
persistent. Also, it allows me to identify the time at which the differentials started to
widen/narrow. Second, I analyze the impact of the two waves of expansion of modern
schooling in the 1930s and 1950s on the inter-religious literacy gap using both data
sources. I am not able to examine the impact of the 1910s expansion since the cohort
born before 1910 is too old to observe in the 1986 census. Also, since the 1986
individual- level census data is not suitable to study the occupational gap for the cohort
affected by the 1950s expansion as members of this cohort were still too young in 1986, I
will use, in a future extension of this research, the 1996 individual-level census data in
order to analyze the impact of the 1950s expansion on the occupational gap.
116

3.4.1. Evolution of Inter-Religious Differentials in Educational and Occupational
Outcomes Over the Twentieth Century

A. Village/Urban Quarter- Level Census Data (1897-1986):
I pool the village/quarter- level census data from all years and run the following OLS
regression (1):

=

+

+ !"
"ℎ

# + !"
"ℎ

× 8

+ *"

>
+ *"

× 8

?

+

Where: y
it
is the outcome of village/quarter i in the census year t. α
i
is village/quarter
fixed effects, β
t
is census year fixed effects, nmuslimshare
it
is the percentage of non-
Muslims in village/quarter i in census year t. T
t
is a full set of dummies for census years.
Controls
it
is a set of controls in village/quarter i in census year t. Outcomes of interest are
percentage of males who are able to read and write, and percentage of males who are
working in non-agricultural activities. Controls include log (population), percentage of
females, percentage of population under 10 years, percentage of population 60 years and
above, and percentage of foreigners. I run the regression with village/quarter fixed
effects, with province fixed effects, and without fixed effects (pooled OLS).
The results are shown in Figures 3.4 and 3.5. These figures show the coefficient δ
t

over time, i.e. the evolution of the inter-religious differentials in literacy and occupational
distribution over the twentieth century with and without village/quarter fixed effects. For
each outcome of interest, I run regression (1) with an increasing set of available controls,
which results in limiting the time period as more controls are added. I do not show the
results from the province fixed effects regressions but they are qualitatively similar.
117

Figure 3.4 shows that literacy gap between non-Muslims and Muslims declined after
1937. In 1897, an increase in the percentage of non- Muslims by one percentage point is
correlated, on average, with an increase in the percentage of literate males by 21.5
percentage points. In 1986, the coefficient switched signs to become negative: A one
percentage point increase in the percentage of non- Muslims is correlated with a decline
in the percentage of literate males by 4.6 percentage points. The evolution of the religious
differentials is qualitatively robust to the addition of the aforementioned controls. Figure
3.5 shows that non-agricultural activity gap declined after 1960. A one percentage point
increase in the percentage of non-Muslims is correlated, on average, with an increase of
34.5 percentage points in 1897, but with a decrease of 12.9 percentage points in 1986.
To summarize, the evidence from the village/quarter- level data suggests that there
was a decline in the inter-religious differentials in educational and occupational outcomes
around mid-twentieth century. I will now turn to the evidence from the individual-level
1986 data.
B. The 1986 Individual- Level 10 percent Census Sample:
While the previous analysis suggests a decline of the educational and occupational
gaps between Muslims and non-Muslims over the twentieth century, the fact that the data
are aggregated at the village/quarter- level might reduce the confidence in the results.
Also, since I am only able to observe male literacy rate and share of males working in
non-agricultural activities in the village/quarter- level data, the analysis is constrained.
Ideally, I want to use other measures of educational attainment (such as years of
schooling) and of occupational attainment (such as being in a white collar occupation).
118

The occupational outcome, in particular, requires a better measure because the share of
males working outside the agriculture sector is actually a measure of economic activity
rather than occupation.
The 1986 individual-level census sample offers a complementary way to analyze the
evolution of the educational and occupational gaps between religious groups across
cohorts. The analysis is constrained by the fact that in 1986 I am only observing the
survivors who chose to remain in Egypt. Nevertheless, this problem is perhaps
concerning for the older cohorts but not so much for the younger ones. Moreover, the two
pieces of evidence together (village/quarter- level data and 1986 individual- level data)
may give a complete and convincing picture. The first piece gives the complete picture at
the time of the censuses at an aggregated level, while the second piece gives a more
detailed snapshot on the survivors observed in 1986.
To analyze the inter-religious differentials across cohorts, I estimate the following
regression (2):

=

+

+ #*ℎ
"

+ *ℎ
"

× 8

+

Where: y
idt
is the outcome of interest for individual i born in district d in decennial
cohort t, α
d
is district of birth fixed effects, β
t
is cohort of birth fixed effects, christian
i
is
an indicator whether individual i is Christian, T
t
is a set of dummies for cohort of birth,
and ε
idt
is the disturbance term. Outcomes of interest are an index for literacy, years of
schooling, an index for working in the non-agricultural sector, and an index for working
in a white-collar occupation. The sample is restricted to males who are at least 10 years
119

old in 1986 when studying literacy and to males who are at least 25 years old in 1986
when analyzing years of schooling and occupational outcomes.
The results are shown in figures 3.6-3.9, which show the evolution of δ
t
across
cohorts. Figure 3.6 shows an accelerated convergence in literacy starting from the cohort
born in (1960-69). Figure 3.7, however, suggests persistence in the years of schooling
gap for the cohort born in (1950-59). Also, there is no evidence for convergence in the
occupational outcomes up to the decennial cohort born in (1950-59).
To sum up, the analysis of the evolution of the inter-religious differences in
educational and occupational outcomes suggests that there was persistence in the gaps in
the first half of the twentieth century and it was only in the second half of the century that
Muslims started to catch up with non-Muslims at least with respect to literacy. In the
future, more evidence is needed from the later 1996 census to analyze the evolution of the
occupational gap for the younger cohort born in (1960-69).
3.4.2 Impact of Expansion of Modern Schooling on Inter-Religious Literacy Gap
In this section, I analyze the impact of the two waves of expansion of modern
schools that occurred in (1927-36) and (1948-59) on the inter-religious literacy gap. Two
notes are in order: First, I do not analyze the impact of the (1907-21) expansion because
the 1986 census sample does not allow this analysis, since the individuals born before
1910, who are still alive in 1986, are more likely to be a highly selected sample of the
cohort’s population. Second, In this chapter I am analyzing the impact of modern
schooling on literacy gap. In a future extension of this research, I will analyze the impact
120

on years of schooling and occupational outcomes using the 1996 individual-level census
sample along with the village/quarter- level data (1897-1986).
A. Evidence from Village/Urban Quarter- Level Data:
To evaluate the expansion programs, I start by matching the village/quarter- level
census data with the schools data at the district- or province- level. I adjust the local
supply of schools by population size. To evaluate the 1927-36 expansion, I restrict my
analysis to the 1927 and 1947 censuses, and to evaluate the 1948-59 expansion, I restrict
my analysis to the 1947 and 1976 censuses. I, hence, allow for some lag for the effect of
school expansion on literacy to be realized. The analysis is carried out by running the
following pooled OLS regression (3):

=
1
+
2
+ !"
"ℎ

3
+ !"
"ℎ

× 8

4
+ "'ℎ"

×
8

5
+ !"
"ℎ × "'ℎ" × 8

6
+ *"

7
+ *"

× 8

8
+

(3)

Where: is the percentage of males who are able to read and write in village i in
district d in census year t, is census year fixed effects, is the lagged supply
of modern schools (per 10,000 individuals) in the district, and is a dummy for the
census year after the expansion program. In this strategy, captures the base effect of the
lagged local supply of schools (lagged for the post-expansion census means that it is
measured at the time of the expansion) on male literacy rate in the first census after the
expansion took place. On the other hand, captures the added impact on male literacy
rate when the lagged supply of schools is interacted with the percentage of non-Muslims
121

in the village or urban quarter. In other words, the two coefficients can be interpreted as
measuring the impact of the program on Muslims and non-Muslims respectively.
Table 3.4 shows the results for the two expansion programs. The 1927-36 expansion
seems to have had a positive but insignificant effect on Muslims, and an additional
positive and significant effect on non-Muslims. This can be interpreted by the fact that
the program was an expansion of non-free modern schools. One would expect that non-
Muslims were poised to benefit more from this expansion due to their higher
socioeconomic status that allowed them to pay for the tuition fees. The 1948-59
expansion, on the other hand, seems to have had a positive and significant effect on
Muslims, and an additional positive but insignificant effect on non-Muslims. This is
consistent with the interpretation that this expansion was accompanied by free schooling.
Muslims and non-Muslims, hence, may have benefited equally from this expansion. To
sum up, according to this piece of evidence, while the first expansion led to an
aggravation of the literacy gap, the second expansion benefited Muslims and Christians
equally.
B. Evidence from Individual- Level 1986 Data:
I complement the above analysis by evidence from the individual- level data from
the 1986 census. I merged the 1986 data with the schools information at the district of
birth- or province of birth- level. The local supply of schools from a given schooling
report year was matched with the individual- level data based on the cohort of birth of the
individual. I focus only on the cohorts that were born before or after the expansion in
modern schooling. In particular, I run the following regression (4):
122

=
1

+
2
+
3
*ℎ
"

+ *ℎ
"

× 8

4
+ "'ℎ

× 8

5
+ *ℎ
"

× "'ℎ

× 8

6
+

Where: is the literacy index for an individual i born in district d in cohort t.
is district of birth fixed effects. is cohort of birth fixed effects. is a dummy for the
“young” cohort who benefited from the expansion. is the local supply of modern
schools in the district of birth at the cohort of birth of the individual. has a direct
interpretation as the base effect of the expansion of modern schooling on “young”
Muslims. measures the additional effect of the expansion on “young” Christians.
Table 3.5 shows the results for the two expansion programs. The 1927-36 expansion
seems to have had a positive (but insignificant) effect on Muslims, and an additional
negative (but insignificant) effect on Christians. Hence, the results suggest that this
expansion did not have a significant impact on the inter-religious literacy gap (while the
evidence from the village/quarter- level data suggested that there was an aggravation of
the gap). On the other hand, the 1948-59 expansion seems to have been much more
successful in closing the literacy gap. The expansion had a positive effect on Muslims but
an additional negative and highly significant effect on Christians.
3.5 Conclusion
Employing new data sources from Egypt, one of the largest Middle Eastern
countries, I analyzed the evolution of the educational and occupational differences
between Muslims, Christians, and Jews over the course of the twentieth century, and the
impact of the expansion of modern schooling on these differences.
123

I find that while the educational differences between Muslims and non-Muslims
seem to have declined in the second half of the twentieth century, the occupational
differences are perhaps still persistent. The expansion of modern schooling in the first
half of the twentieth century did not help reduce the literacy gap between Muslims and
non-Muslims, perhaps because these modern schools imposed high tuition fees. On the
other hand, the expansion of free modern schooling in the aftermath of the 1952 military
coup was perhaps more successful in reducing the literacy gap between Muslims and
non-Muslims. More research has to be done to analyze the impact of these expansion
programs on the occupational differences.
The findings suggest that public policy in education can help reduce the religious
differentials in educational outcomes if the design of the policy takes into consideration
the underlying reasons behind the differentials. An expansion of non-free schooling is
less likely to reduce religious gaps since the religious groups with the higher
socioeconomic status may benefit more from such expansion due to their ability to pay
for the high tuition fees. An effective expansion of modern schooling has to induce
higher demand for education by increasing the expected return to education and reducing
its cost so that the expected returns are likely to offset the costs for the dispriviliged
religious group.

124

Figure 3.1: Mean Literacy Rates of Muslims and Christians across
Decennial Birth Cohorts (1910-70)
(Males who are at least 10 years old in 1986 Census)

0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1910 1920 1930 1940 1950 1960 1970
Birth Cohort
Mean Christian
Literacy
Mean Muslim Literacy

Figure 3.2: Percentage of Muslims and Christians Working in White-
Collar Occupations across Decennial Birth Cohorts (1910-50)
(Males who are at least 27 years old in 1986 Census)

0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1910 1920 1930 1940 1950
Birth Cohort
Mean Christian White
Collar
Mean Muslim White
Collar

125

Figure 3.3: Evolution of Supply of Elementary Schools (kuttabs) and
Modern Schools in the Period 1907-69

0
1
2
3
4
5
6
7
8
9
1907 1921 1927 1936 1948 1959 1969
Year
Mean Kuttabs (per
10,000 individuals)
Mean Modern Schools
(per 10,000
individuals)

126

Figure 3.4: Regression of % Males Who Are Able to Read and Write on
% Non-Muslims
(Village/Urban Quarter- Level Census Data 1897-1986)

(Without Village/Quarter Fixed Effects: Pooled OLS)
-0.15
-0.1
-0.05
0
0.05
0.1
0.15
0.2
0.25
1897 1907 1917 1927 1937 1947 1960 1976 1986
Coefficient
Year
Coefficient (Basic Controls:
log(pop), % females)
Coefficient (Adding controls for
% under 10 and % 60 and above)
Coefficient (Adding control for
% foreigners)

(With Village/Quarter- Fixed Effects: Panel Regression)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
1897 1907 1917 1927 1937 1947 1960 1976 1986
Coefficient
Year
Coefficient (Basic Controls:
log(population) and %
females)
Coefficient (Adding controls
for % under 10 and % 60 and
above)

127

Figure 3.5: Regression of % Males Working in the Non-Agricultural
Sector on % Non-Muslims
(Village/Urban Quarter- Level Data 1927-1986)

(Without Village/Quarter Fixed Effects: Pooled OLS)
-0.2
-0.1
0
0.1
0.2
0.3
0.4
1927 1937 1947 1960 1976 1986
Coefficient
Year

(With Village/Quarter Fixed Effects: Panel Regression)
0
0.05
0.1
0.15
0.2
0.25
0.3
1927 1937 1947 1960 1976 1986
Coefficient
Year

128

Figure 3.6: Religion and Literacy across Birth Cohorts (1910-70)
(Individual- Level Data 1986 Census Sample)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1910 1920 1930 1940 1950 1960 1970
Coefficient
Year
Marginal Effect for
Christians
Marginal Effect for
Muslims

This graph is based on the regression of literacy dummy on Christian dummy, birth cohort dummies,
district of birth dummies, and interactions of birth cohort dummies and Christian dummy.

Figure 3.7: Religion and Years of Schooling across Birth Cohorts (1910-
50) (Individual- Level Data 1986 Census Sample)
0
1
2
3
4
5
6
7
8
9
1910 1920 1930 1940 1950
Coefficient
Year
Marginal Effect for
Christians
Marginal Effect for
Muslims

This graph is based on the regression of years of schooling on Christian dummy, birth cohort dummies,
district of birth dummies, and interactions of birth cohort dummies and Christian dummy.
129

Figure 3.8: Religion and Employment in the Non-Agricultural Sector
across Birth Cohorts (1910-50)
(Individual-Level Data 1986 Census Sample)
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
1910 1920 1930 1940 1950
Coefficient
Year
Marginal Effect for
Christians
Marginal Effect for
Muslims

This graph is based on the regression of non-Agricultural sector dummy on Christian dummy, birth cohort
dummies, district of birth dummies, and interactions of birth cohort dummies and Christian dummy.

Figure 3.9: Religion and Employment in White-Collar Occupations
across Birth Cohorts (1910-60)
(Individual-Level Data 1986 Census Sample)
0
0.05
0.1
0.15
0.2
0.25
0.3
0.35
0.4
0.45
0.5
1910 1920 1930 1940 1950
Coefficient
Year
Marginal
Effect for
Christians
Marginal
Effect for
Muslims

This graph is based on the regression of white collar dummy on Christian dummy, birth cohort dummies,
district of birth dummies, and interactions of birth cohort dummies and Christian dummy
130

Table 3.1: Descriptive Statistics Across Years: Village/Urban Quarter-
Level Census Data (1897-1986)

Variable
1897 1907 1917 1927 1937 1947 1960 1976 1986
% Non-Muslims 0.07
(0.16)
0.07
(0.15)
0.07
(0.15)
0.07
(0.15)
0.07
(0.14)
0.06
(0.14)
0.06
(0.13)
0.05
(0.12)
0.05
(0.12)
% Christians
0.07
(0.16)
0.07
(0.14)
0.07
(0.14)
0.06
(0.13)
0.06
(0.13)
0.05
(0.12)
0.05
(0.12)
% Coptic Christians
0.06
(0.14)
0.06
(0.13)
0.06
(0.13)
0.06
(0.13)
% Non-Coptic
Christians
0.01
(0.07)
0.01
(0.06)
% Jews
0.00
(0.02)
0.00
(0.03)
0.00
(0.03)
0.00
(0.00)
0.00
(0.00)
0.00
(0.00)
% Males Able to Read
and Write
0.08
(0.08)
0.08
(0.07)
0.11
(0.09)
0.17
(0.11)
0.21
(0.13)
0.24
(0.13)
0.24
(0.12)
0.35
(0.15)
0.40
(0.13)
% Males With At
Least High School

0.05
(0.07)
0.07
(0.08)
0.17
(0.10)
% Males With At
Least University
0.01
(0.02)
0.02
(0.04)
0.03
(0.04)
% Males in Non-
Agriculture Sector
0.30
(0.26)
0.31
(0.25)
0.34
(0.26)
0.27
(0.28)
0.33
(0.29)
0.42
(0.26)
% Males in White
Collar Occupations

0.04
(0.06)
0.07
(0.06)
0.10
(0.07)
% Foreigners
0.01
(0.07)
0.01
(0.06)
0.01
(0.05)
0.00
(0.03)
% Population Under 5

0.14
(0.02)
0.13
(0.03)
0.13
(0.02)
0.16
(0.02)
0.14
(0.03)
0.16
(0.03)
% Population Under 10

0.27
(0.03)
0.27
(0.04)
0.26
(0.03)
0.30
(0.03)
0.28
(0.04)
0.29
(0.05)
% Population 60 and
Above

0.07
(0.02)
0.06
(0.02)
0.06
(0.02)
0.07
(0.02)
0.07
(0.02)
0.06
(0.02)
N
4,044 3,828 3,839 3,903 4,200 4,342 4,682 4,791 4,910

Frontier provinces are excluded. Standard deviation is between parentheses.

131

Table 3.2: Descriptive Statistics Across Regions: Village/Urban
Quarter- Level Census Data (1897-1986)

1897
Variable Urban
Provinces
Lower
Egypt
Northern
Upper
Egypt
Middle
Upper
Egypt
Southern
Upper
Egypt
% Non-Muslims 0.147 0.028 0.049 0.183 0.040
% Christians 0.132 0.028 0.049 0.183 0.040
% Coptic Christians 0.027 0.027 0.049 0.182 0.040
% Non-Coptic Christians 0.105 0.001 0.000 0.000 0.001
% Jews 0.014 0.000 0.000 0.000 0.000
% Males Able to Read and Write 0.195 0.076 0.054 0.047 0.040
% Foreigners 0.098 0.001 0.000 0.000 0.001
N
432 2,184 462 738 228
1986
Variable Urban
Provinces
Lower
Egypt
Northern
Upper
Egypt
Middle
Upper
Egypt
Southern
Upper
Egypt
% Non-Muslims 0.061 0.014 0.033 0.157 0.043
% Christians 0.057 0.014 0.032 0.157 0.043
% Jews 0.000 0.000 0.000 0.000 0.000
% Males Able to Read and Write 0.554 0.423 0.335 0.317 0.380
% Males With At Least High School 0.284 0.169 0.125 0.114 0.148
% Males With At Least University 0.085 0.026 0.019 0.012 0.014
% Males in Non-Agriculture Sector 0.840 0.393 0.378 0.270 0.457
% Males in White Collar Occupations 0.170 0.100 0.081 0.070 0.102
N
547 2,539 610 908 306

132

Table 3.3: Descriptive Statistics: Individual-Level 1986 10 percent
Census Sample

Variable Mean N
% Christians (in the entire sample)
6.19 2,338,748
% Born 1910-19
1.94 2,338,748
% Born 1920-29
4.6 2,338,748
% Born 1930-39
6.61 2,338,748
% Born 1940-49
9.48 2,338,748
% Born 1950-59
13.56 2,338,748
% Born 1960-69
19.15 2,338,748
% Born 1970-79
23.44 2,338,748
% Born 1980-86
21.21 2,338,748
% Literate (Out of males 10 years or older)
65.03 1,680,397
Average years of schooling (For males 25 years or older)
4.81 932,270
% With only primary school degree (Out of males 25 years
or older)
1.86 932,270
% With only preparatory school degree (Out of males 25
years or older)
3.11 932,270
% With only high school degree (Out of males 25 years or
older)
17.06 932,270
% With only university degree (Out of males 25 years or
older)
9.28 932,270
% With scientific majors (Out of males 25 years or older
with at least university degree)
48.88 89,374
% Working in non-agricultural sector (Out of males 25
years or older)
68.2 761,172
% Working in white-collar occupations (Out of males 25
years or older)
25.09 761,657

Notes: 1. Sample is restricted to males born on or after 1910 inside Egypt and who have non-missing values
for age, education, and district of birth.
2. White-collar occupations include personnel in technical and scientific occupations, managers and
business owners, personnel in administrative occupations, managers of wholesale and retailer activities,
managers of hotels, restaurants, and other services sector, and police officers.
3. Scientific majors include medicine, engineering, science, and agriculture majors.

133

Table 3.4: Impact of the Expansion of Modern Schooling on Inter-
Religious Literacy Gap
(Evidence from Village/Urban Quarter- Level Data)
(Dependent variable: % Males who are able to read and write)

1927-36
Expansion
1948-59
Expansion
% Non-Muslims 0.141***
(0.028)
0.059*
(0.033)
% Non-Muslims × Post-Expansion
Census Dummy
-0.120***
(0.020)
-0.292
(0.186)
Lagged Supply of Modern
Schools × Post-Expansion Census
Dummy
0.002
(0.001)
0.037***
(0.006)
% Non-Muslims × Lagged Supply
of Modern Schools × Post-
Expansion Census Dummy
0.013**
(0.007)
0.031
(0.053)
R- Squared 0.237 0.43
N 7542 8396

Regression in the first column includes only villages and quarters from 1927 and 1947 censuses.
Regression in the second column includes only villages and quarters from 1947 and 1976 censuses.
Controls include post-expansion census dummy, log (population), % females, % under 10 years, and % 60
years and above and their interactions with the post-expansion census dummy. Standard errors are in
parentheses and are clustered at the district level.

134

Table 3.5: Impact of the Expansion of Modern Schooling on Inter-
Religious Literacy Gap
(Evidence from Individual- Level 1986 Census Sample)
(Dependent variable: Literacy Index)

1927-36
Expansion
1948-59
Expansion
Christian 0.158***
(0.010)
0.125***
(0.007)
Young Cohort 0.152***
(0.003)
0.111***
(0.010)
Christians ×
Young Cohort
-0.033***
(0.009)
0.073***
(0.027)
Supply of Modern
Schools × Young
Cohort
0.0004
(0.0003)
0.003
(0.003)
Supply of Modern
Schools ×
Christian ×
Young Cohort
-0.0004
(0.0003)
-0.031***
(0.007)
R- Squared 0.149 0.128
N 224,580 453,461

In the first regression, the sample is restricted to the cohorts born in (1912-22) and (1932-42). In the second
regression, the sample is restricted to the cohorts born in (1952-62) and (1932-42). Robust standard errors
are in parentheses and are clustered at the village/quarter- level. Controls include district of birth fixed
effects.
135

Chapter 4
On the Road to Heaven: Poll Tax, Religion, and Human
Capital in Medieval and Modern Egypt

4.1 Introduction
Scribes in the Levant and Egypt are Christians.

Al-Muqaddasi, Muslim historian and geographer, tenth century
1

In 1996 Egypt, Copts (Egyptian Christians) (6 percent of the population) had better
economic outcomes than Muslims. 41 percent of adult active male Copts worked in
white-collar occupations, and 64 percent were literate, as opposed to 30 percent and 61
percent respectively among Muslims.
2
The Coptic-Muslim gap persisted throughout the
twentieth century despite the expansion of modern schooling that reduced the gap, and
the out-migration of high-skilled Copts starting from the 1960s (See Chapter 3). Egypt’s
censuses of 1848 and 1868 reveal that the gap was even higher in the nineteenth century
(See Chapter 2).
3
48 percent of adult active male Copts then worked in white-collar
occupations (mostly scribes), as opposed to 16 percent among Muslims.
4
The

1
Al-Muqaddasi (1877), p. 183.
2
Based on the author’s calculations from the 1996 10 percent census sample published on IPUMS
International. Sample is restricted to males who are at least 25 years old.
3
Based on the author’s calculations from the 1848 and 1868 digitized census samples (See Chapters 1 and
2 for more details about the data source). Sample is restricted to males who are at least 15 years old.
4
The educational gap was also higher. In 1848 Cairo, 51 percent of Coptic male children of 5-14 years of
age were enrolled in Coptic kuttabs (elementary religious schools), which taught arithmetic and geometry
136

phenomenon, the better outcomes of non-Muslims, is observed throughout the Middle
East (Courbage and Fargues 1997, pp. 174-209).
Why did the Coptic-Muslim human capital gap in Egypt emerge?
5
I hypothesize that
the poll tax (jizya) on non-Muslims that was imposed upon the Islamic Conquest of the
then-Coptic Christian Egypt in 640 AD, and remained in effect until 1855, led to the
conversion of poor Copts to Islam to avoid paying the tax, and to the shrinking of Copts
into a better off minority. A testable prediction of the hypothesis is that areas where the
tax was more enforced or where conversion was more responsive to the tax (i.e. with
higher tax elasticity of conversion) witnessed larger conversions to Islam among poor
Copts. These areas should thus have both lower Coptic population share and wider
Coptic-Muslim gap. Medieval institutions of internal-migration-control and skill-
acquisition limited both geographic and occupational mobility, and led to the persistence
of the gap and its spatial pattern over the centuries.
I show that four historical factors predict the variation in Coptic population share,
which is, in turn, inversely related to the size of the gap. These factors are: (1) the eighth
and ninth centuries tax revolts that were ignited by stricter tax enforcement, (2) the Arab
immigration waves to Egypt in the seventh to ninth centuries that Islamized the local
elites, who were responsible for tax collection, and who, depending on their religion,

besides Arabic orthography (Heyworth-Dunne 1938, p. 85). Muslim kuttabs, focusing solely on
memorizing Quran, enrolled only 34 percent.
5
I focus on the Coptic-Muslim gap since Copts constituted 94 percent of non-Muslims in 1848/68. Other
non-Muslim minorities include Jews (1 percent) and non-Coptic Christians (5 percent), such as Ruum
(Ottoman Greeks), Armenians, Greeks (from the independent Greek state), and Levantines. These groups
have better economic outcomes than both Copts and Muslims. Unlike Copts, they are urban groups.
137

exercised more (or less) lenience towards Coptic taxpayers, (3) the Coptic churches and
monasteries in the twelfth and fifteenth centuries, which provided aid to poor Copts, and
(4) the attachment to Christianity measured by the route of the Holy Family in their
biblical flight to Egypt. While tax revolts and Arab immigration measure the variation in
tax enforcement, Coptic institutions and the route of the Holy Family measure the
variation in tax elasticity of conversion. I draw on a wide range of qualitative evidence to
support these findings.
I test the hypothesis using the earliest comprehensive microdata source from Egypt,
two national individual-level samples of the 1848 and 1868 Egyptian censuses that I
digitized from the original manuscripts. I also make use of the village/urban quarter-level
1897 census data, which have the full population enumeration. Historical evidence,
however, suggests that the gap emerged by the twelfth century. I choose the nineteenth
century censuses rather than the most recent ones because the earlier data provide a
snapshot of the Egyptian population before the expansion of education, urbanization, and
Coptic out-migration that may have altered the gap and its spatial pattern during the
twentieth century. I also created datasets on medieval Coptic tax revolts, post-Conquest
Arab immigration waves, medieval Coptic churches and monasteries, and the route of the
Holy Family’s journey in Egypt, using various historical sources.
6

6
I use Morimoto (1981) and Mikhail (2004) for the information on the eighth and ninth centuries tax
revolts, Al-Barry (1992) for the dataset on Arab immigration waves, the medieval sources Abul-Makarim
(1999) and Al-Maqrizi (2002) for Coptic churches and monasteries in the twelfth and fifteenth centuries,
and Anba-Bishoy (1999) and Gabra (2001) for the sites visited by the Holy Family during their biblical
flight to Egypt.
138

My explanation of the Coptic-Muslim gap differs from that of other scholars. First, a
tradition in Middle Eastern economic history (e.g. Issawi 1981; Kuran 2004) postulated
that the better economic outcomes of non-Muslim minorities did not exist before the
nineteenth century. It has thus been suggested that non-Muslims’ privilege was due to the
rise of modern European influence in the Middle East that favored local non-Muslims
(Issawi 1981), or to their adoption of European legal structures (Kuran 2004).
Nonetheless, a large body of evidence suggests that Copts’ quasi-monopoly of white-
collar occupations, which accounts for the bulk of the gap in the 1848/68 censuses,
emerged by the twelfth century (See subsection 4.6.1). Second, one may argue following
Kuznets (1960) that Copts, as a minority, attempted to preserve their religious identity by
specializing in administrative occupations in which they had built a tradition. Although
this may seem appealing in explaining the persistence of Coptic monopoly of
administrative jobs once Copts became a minority, it does not explain how Copts, who
initially constituted almost the entire population, made the occupational transition. Third,
the white-collar bias of the Jewish occupational distribution is often explained by the fact
that Jews were banned from specific occupations such as farming (Abrahams 1896, pp.
211-50). But Copts were not banned from farming and, unlike Jews, were not an “urban”
population. In 1848/68, 33 percent of adult active male Copts were farmers.
7
Finally,
Botticinni and Eckstein (2005) argued that the literacy requirement under Rabbinic
Judaism led, via positive selection, to the Jewish higher human capital. Yet, no literacy

7
The fact that, in 1848-68, Copts were excluded from top state positions in the military, the police, the
judiciary, and the ulama, which were all high-status jobs, does not explain the positive Coptic-Muslim gap.
The opposite scenario, banning Muslims from specific “Coptic” occupations (such as making, selling, or
serving alcohol) also existed. Yet, the share of these occupations is too limited and their status is not high.
139

requirement existed under Coptic Christianity, and illiteracy among adult male Copts in
1986 was as high as 34 percent.
The chapter contributes to the economics literature on the lasting impact of
institutional changes on economic outcomes (e.g. Acemoglu, Johnson, and Robinson
2001; Acemoglu et al. forthcoming). I show in this chapter that the Islamic Conquest of
Egypt that established the poll tax institution in 640 AD created unintentionally religious
human capital differentials that persisted for over a millennium. It is also related to the
literature on the impact of religion on economic outcomes (Botticinni and Eckstein 2002
and 2005; Barro and McCleary 2003 and 2005; Becker and Woessmann 2009; Boppart et
al. 2008; Boorooah and Iyer 2005; Chaudhary and Rubin 2009). I provide a new
explanation for the correlation between religion and economic outcomes based on the
historical self-selection into the religion, which is caused by irreligious factors (poll tax).
Finally, the effect of the poll tax on the conversion of Copts to Islam and the monopoly of
Copts over administrative jobs in medieval and modern Egypt are both old topics in the
historical tradition (On poll tax and conversion: e.g. John of Nikiu 1916; Severus 1907;
Dennett 1950; Morimoto 1981. On Coptic scribes: e.g. Al-Maqrizi 2002; Al-Muqqadasi
1877; Tagher 1998; Sheikho 1987; Samir 1996; Amer 2000). Yet, this research is, to the
best of my knowledge, the first to argue that the poll tax led to selected conversion of
Copts to Islam (rather than systematic conversion), and that that, in turn, generated the
human capital gap. My key point here is shifting the attention from identifying the
reason(s) behind the Islamization of Egypt per se, a vast topic in the historical literature
that generated fierce debates, into identifying the reason(s) behind the emergence of the
140

Coptic-Muslim human capital gap. This gap is rarely emphasized in the historical
literature although the Coptic dominance of administrative occupations is rarely disputed.
The new data sources that I collected in support of the hypothesis may thus shed a new
light on both the Coptic-Muslim gap and the historical process of Islamization of Egypt.
The rest of the chapter is organized as follows. Section 4.2 provides historical
background on the Christianization and Islamization of Egypt. I discuss the data in
section 4.3. Section 4.4 and 4.5 introduce the empirical analysis and the results. I discuss
the mechanisms through which the hypothesis operates in section 4.6. Finally, section 4.7
concludes.
4.2 Historical Background: Christianization and Islamization of Egypt
Christianity has a long history in Egypt, reaching back to the 1
st
century, and the
Church of Alexandria was a major theological center since the 2
nd
century (Bowman
1989, pp. 191-202; Roberts 1979, pp. 1-26).
8,9
Both the translation of biblical texts into
Coptic in the 3
rd
century and Constantine’s Edict of Toleration in 331 accelerated the
Christianization of Egypt. Thus, by the end of the 4
th
century, incidents of persecution of
pagans by the now dominant Christians were documented,
10
and the last pocket of

8
According to Coptic traditions, Egypt knew Christianity through the biblical flight of the Holy Family to
Egypt, and then through Saint Mark, the author of the Gospel of Mark, who is believed to have established
the Church of Alexandria (Atiya 2005, pp. 29-37). Apart from this tradition, both Bowman (1989, p. 191)
and Roberts (1979, p. 1) suggest that the presence of a large Hellenized Jewish community in Alexandria
may have encouraged the choice of Egypt by early Christian preachers as their first station in the
propagation of Christianity after Palestine.
9
The first evidence on Christianization comes from the 3
rd
century: a document evidencing the persecution
of Egyptian Christians. Decius, the Roman emperor, ordered that every individual had to declare that he
was not a practicing Christian, and the evidence is one of these declarations (Bowman 1989, p. 191).
10
This includes the murder of the Greek Neo-Platonist philosopher Hypatia by a Christian mob in 415
(Bowman 1989, p. 207). Interestingly, paganism survived longer among the Hellenized elite.
141

paganism was Christianized in the mid-6
th
century (Bowman 1989, p. 192). The Church
of Alexandria, followed by the Coptic (Egyptian) masses, separated from the (ruling)
Byzantine Church on the grounds of a theological debate at the Council of Chalcedon in
451 (Atiya 2005, pp. 71-6; Tagher 1998, pp. 1-7).
11
Greeks and Hellenized Egyptians
remained loyal to the Byzantine Church forming a competing church, the Melkite Church
of Alexandria.
12
Condemned as “heretics,” Copts suffered persecution under the
Byzantines until the Islamic Conquest (Bowman 1989, p. 198; Atiya 2005, pp. 87-99),
which created animosity between Copts and Melkites throughout the Middle Ages.
Hence, when Muslims conquered Egypt in 640, Copts constituted almost all the
population, with Melkites and Jews forming two small minorities (Tagher 1998, p. 4;
Lane-Poole 1969, p. 2; Wilfong 1998).
13
Over the centuries, Egypt was gradually
Islamized, and Copts (Egyptian anti-Chalcedonian Christians) fell to 6 percent of the
population in 1848. Two historical processes account for the Islamization of Egypt:
Conversion of Copts to Islam and Arab (Muslim) immigration to Egypt. Demographic
evidence in Russell (1958) suggests that the size of Arab immigration cannot, by itself,

11
The “Great Schism” that separated the Byzantine and Roman Churches occurred later, in 1054. The
Chalcedonian theological controversy was on the nature of the Christ. The Coptic Church of Alexandria
rejected the Chalcedonian decree and advocated for the “Monophysite” doctrine, i.e. that the Christ has a
single inseparable nature. The Melkite Church, a pro-Chalcedonian Church, believed in the two natures
(divine and human) of the Christ (Atiya 2005, 87-9), and it later followed the Greek Orthodox Church.
12
According to Atiya (2005, pp. 21-2, 88), the word “Melkite” originated from the Syriac word for “King,”
and was the name used by Copts to describe their opponents as followers of the Byzantine Emperor or
“King.” The word “Copt,” on the other hand, originated from the Greek word for Egypt.
13
There were other minor Christian factions in Egypt that were later assimilated into the Coptic or Melkite
Churches. These included Melitians, Arians, Manicheans, Eutychians, Acephali, Barsanuphians, and
Julianists (See Mikhail 2004, p. 46 for more details).
142

account for Islamization,
14
although inter-marriage between Arab immigrants and Copts
may have acted as a “multiplier” of the base immigration effect.
15
Importantly, there was
no significant Coptic emigration from Egypt, or “replacement” of Copts by Arab
immigrants. Copts, because of their anti-Chalcedonian denomination that differed from
both Eastern and Western Churches, were a highly isolated and immobile group that was
first “discovered” by Europeans in the fifteenth century (Hamilton 2006, pp. 1-5). Large-
scale Coptic emigration from Egypt only started in the 1960s.
In the absence of statistics, Wilfong (1998) points out that it is not possible to
identify the point at which Copts became a minority. An old historical tradition (e.g. Al-
Maqrizi 2002; Lane-Poole 1969; Dennett 1950; Mikhail 2004) argued that Islamization
(i.e. Copts declining to a minority) took place in the early period (640-1250) especially
after the suppression of the eighth and ninth centuries tax revolts. A more recent tradition
(e.g. Wiet 1927; Little 1976; Al-Leithy 2005; Werthmuller 2010) argued that the
“deathblow” to Christianity only occurred in the Mamluk period (1250-1517) as various
pressures by the state triggered a wave of mass conversion to Islam among Copts. A third
school attempted to use quantitative methods. Bulliet (1979) used lineages of a sample of
prominent individuals in medieval narratives to identify the point at which an individual’s
ancestors converted to Islam and thus adopted an Arabic name. He found that

14
The population of the Arab peninsula in 600 AD is estimated at around 1 million, while the population of
Egypt was around 2.7 millions (Russell 1958, p. 89). Thus, had the entire Arab population migrated to
Egypt, it would not have sufficed, by itself, to account for the almost complete Islamization of the country.
15
Fertility and mortality differentials between Arabs and Copts in medieval Egypt are also a possibility.
Muslims in twentieth century Egypt had higher crude birth and death rates than Christians (Courbage and
Fargues 1997, pp. 199-200).
143

conversions peaked in the ninth century.
16
Courbage and Fargues (1997), following
Russell (1958), used a time series of total poll tax receipts (from medieval narratives) to
estimate the number of non-Muslim taxpayers assuming a fixed nominal tax per capita
over time (which is consistent with the historical tradition). They found a sharp decline in
the receipts before 800, thus suggesting that Copts became a minority by then.
4.3 Data
The empirical analysis is based on two national samples (80,000 observations each)
of the 1848 and 1868 Egyptian censuses that I digitized from the original manuscripts at
the National Archives of Egypt and two additional oversamples of non-Muslims in Cairo
in each year (4,000 observations each) (See Chapters 1 and 2). These censuses are the
earliest microdata source on the Coptic-Muslim gap in Egypt. I pooled the samples and
restricted the analysis to Muslim and Coptic adult active males who are at least 15 years
old in 1848 (or 1868). I also use the village/urban quarter- level 1897 census data, which
have the full enumeration, to measure the population and the Coptic population share in
each geographic unit (See Chapter 3). In the absence of income in the 1848/68 dataset, I
define the religious economic gap in terms of the Social Status Index (henceforth, SSI), a
measure of occupational attainment assigning values from 1 to 12 for each occupational
title, with 12 being the highest social status.
17
Table 4.1 shows that Copts have higher
mean SSI than Muslims, and this is confirmed by figure 4.1, which shows that the Coptic
SSI distribution is more skewed to the right. Table 4.1 also shows that the religious

16
This is, however, subject to an obvious caveat: the representativeness of the biographies’ sample.
17
See Appendix C for details on the SSI construction.
144

distribution at both the national and regional levels and the regional distribution of Copts
are similar across 1848/68 and 1897, thus suggesting their stability over the second half
of the nineteenth century, and confirming the national representativeness of the 1848/68
sample.
I drew on various sources to measure three historical factors employed in the
empirical analysis: Medieval Arab immigration to Egypt, Coptic churches and
monasteries, and the route of the Holy Family.
18
First, I constructed an index at the 1897
district-level that takes the value of one if the district had received any Arab immigration
wave in the seventh to ninth centuries. The information is based on Al-Barry (1992),
who, in turn, draws on Arabic medieval narratives. The author traces, for almost all the
immigrating tribes, their destination in Egypt mostly at the kura-level, the administrative
division that existed in 640-1036. Drawing on Ramzi (1994) and Tousson (1926), I
matched the kuras to the 1897 districts, based on the location of the chef-lieu of each
kura.
Second, I measured the number of churches and monasteries in the Middle Ages at
the 1897 urban district- and rural village- level, drawing on two medieval sources: (1)
Abul-Makarim (1999) provides a comprehensive list of Christian religious
establishments, Coptic, Melkite, and Armenian, that existed in Egypt at the end of the
twelfth century. (2) Al-Maqrizi (2002) includes a less comprehensive list of churches and

18
See Table 4.4 for the data on the first historical factor: tax revolts in the eighth and ninth centuries. See
the Appendix D for more details on the measurement of the other three historical factors.
145

monasteries (both Coptic and Melkite) in the fifteenth century. Both lists are organized
on a geographic basis at the urban street- and rural village-levels.
Finally, as a proxy for the local attachment to Christianity, I constructed an index at
the 1897 urban district- and rural village- level that takes the value of one if the
geographic unit had been visited by the Holy Family during their biblical flight to Egypt,
using the information in Anba-Bishoy (1999) and Gabra (2001). The route of the Holy
Family in Egypt was recorded according to ancient and medieval Coptic local traditions,
which preserved the “miraculous” sites that were touched by the Holy Family (e.g. hand
and footprints of the child Jesus, trees that were believed to have sheltered Mary, and
wells that were believed to be “created” by Jesus).
Table 4.2 shows the descriptive statistics for these datasets. 17 percent of Arab
immigration waves were temporary (Irtiba’) and were thus excluded from the sample.
19

66 percent of the permanent Arab immigration waves were from Qahtan tribes in
Southern Arab Peninsula, and about half of the waves were destined to urban Egypt.
More than 90 percent of churches and monasteries in the twelfth and fifteenth centuries
were Coptic. The percentage of churches and monasteries that were demolished shortly
before the time of the medieval chronicle is particularly large in the fifteenth century (43
percent), because of the so-called “Churches’ Event” under which Muslim masses burned
a large number of churches nationwide (Al-Maqrizi 2002, pp. 1066-76). Interestingly, the
regional distribution of the fifteenth century Coptic churches and monasteries that

19
Irtiba’ comes from the Arabic word for “spring,” and it refers to the post-Conquest tradition that Arab
tribes were moving in the spring season to any Egyptian village of their choice for grazing their animals.
Egyptians (Copts) were required to provide them with food and shelter (Al-Barry 1992, pp. 56-60).
146

survived resembles the Coptic regional distribution in 1897 with the largest Coptic
concentration being in the Nile Valley (Table 4.1). The majority (57 percent) of the sites
visited by the Holy Family are located in the Nile Valley too.
4.4 Empirical Analysis: Coptic Population Share and the Coptic-
Muslim Gap

A testable prediction of the hypothesis that the poll tax institution led to selected
Islamization and thus to the emergence of the religious economic gap is that areas where
the tax was more enforced, or where the demand for Islam was more responsive to the tax
incentive, witnessed larger conversions to Islam among poor Copts. Under geographic
and occupational immobility, these areas should have lower Coptic population share and
wider Coptic-Muslim gap in the nineteenth century. I start the analysis by examining the
validity of the second part of this prediction: Do areas with lower Coptic population share
have larger religious economic gap in the nineteenth century? Using the 1848 and 1868
pooled individual-level samples I estimate the following linear regression:
(1) ε μ φ δ γ β α + + × + + + + =
jk jk ijk jk ijk k ijk
lpop CoptShare Copt CoptShare Copt SSI ) (
Where the SSI of an individual i residing in area j in greater area k depends on the
following regressors: Copt, a dummy variable for being Coptic Christian, CoptShare, the
percentage of Copts in the area of residence j in the 1897 census, and the interaction of
the two terms.
20
The regression also controls for lpop, the logarithm of the population of
the area j in 1897, and greater area k fixed effects. The area j is defined at four levels

20
I use the 1897 census for data on population and Coptic population share because the village-level data of
that census include the full enumeration. The 1848 and 1868 census datasets are individual-level samples.
147

(from smallest to largest geographic division): (1) district in urban provinces and village
in rural provinces, (2) district, (3) province, and (4) region. The greater area k is defined
as the level higher than j: (1) urban province and rural district, (2) province, and (3)
region. If j is defined at the region-level, no fixed effects are included in the regression.
Although, the dependent variable of interest is the difference between the average Coptic
and Muslim SSIs in a geographic unit, I conduct the analysis at the individual-level
because of the data limitation. Since the 1848/68 individual-level samples are only
representative at the province-level in rural Egypt (See Chapter 1), I cannot compute the
Coptic-Muslim SSI gap at the rural village- or even rural district- levels.
21

The results are shown in table 4.3. Regardless of the unit of geography at which
CoptShare and lpop are measured, a Copt has greater SSI than a Muslim by more than
one standard deviation (~ 2 points). An increase in the Coptic population share in the area
of residence has a positive but insignificant impact on the SSI for Muslims, thus implying
no significant spillover effects on Muslims from residing in areas with higher Coptic
concentration. Most importantly, I find that the coefficient on the interaction term is
negative and statistically significant in all regressions: The Coptic-Muslim SSI gap in an
area with high Coptic population share is smaller than in areas with low Coptic
population share. For example, the estimated Coptic-Muslim SSI gap in Middle Nile
Valley, the region with the highest Coptic population share (19 percent), is 0.41. But the

21
I also conducted the analysis using the village/urban quarter- level 1897 census data with the outcome of
interest being the percentage of males who are able to read and write (literacy rate). The 1897 data,
however, do not provide religion-specific literacy rates. I estimated the following regression by
region: literacyrate=α+βCoptshare+γlpop+ε . Similar to my findings in the 1848/68 individual-
level census samples, I found that the magnitude of β is highest in the Nile Delta, which has the lowest
Coptic population share, and lowest in the Middle Nile Valley, with the highest Coptic population share.
148

gap increases to 2.137 in the Nile Delta, where Copts constitute only 2 percent of the
population. This lends support to the second part of the hypothesis: Areas where Copts
are relatively more concentrated tend to have lower Coptic-Muslim economic gap. I now
turn to examine the first part of the hypothesis: What are the historical origins of the
Coptic spatial distribution (variation in Coptic population share) in Egypt?
4.5 Empirical Analysis: Historical Origins of the Coptic Spatial
Distribution

4.5.1 Islamic Poll Tax Institution: Historical Background
My hypothesis that the Islamic poll tax institution led to the “selected” conversion of
poor Copts to Islam and thus to the emergence of the Coptic-Muslim gap builds on a
long-standing historiography. Historians have long emphasized the impact of the poll tax
on the “systematic” conversion of Copts to Islam. The earliest evidence comes from the
seventh century Coptic chronicle of John of Nikiu (1916, p. 201) who stated that “(Arabs)
increased the taxes… And now many of the Egyptians who had been false Christians
denied the holy orthodox faith and lifegiving baptism, and embraced the religion of the
Moslem, the enemies of God.” The ninth century Coptic chronicle of Severus (1910, pp.
116-7) described that, in the early eighth century, “(Arabs) proclaimed that all those, who
would give up their own religion and become Muslims, should be exempted from the poll
tax… By means of this procedure Satan did much harm to many people who gave up their
religion… and they amount to twenty-four thousand persons.” Modern historiography
attempted to reconcile these medieval narratives with the emerging wealth of Arabic
papyri evidence (e.g. Wellhausen 1927; Becker 1903; Dennett 1950; Morimoto 1981).
149

In this subsection, I address several aspects of the poll tax historiography that are
central to my hypothesis, before turning to the empirical evidence. These aspects are: (1)
The tax on non-Muslims was imposed at the individual-level, (2) The tax burden was
higher at lower levels of income, (3) The poor were not exempt from the poll tax, (4)
Muslim converts were exempt from the poll tax, and (5) The real poll tax amount did not
have a trend over the Middle Ages. In what follows, I will discuss each of these points.
Was the tax on non-Muslims indeed a poll tax imposed at the individual-level? From
750 onwards, Islamic rulers and jurists distinguished between the poll tax (jizya),
imposed on non-Muslims on the individual basis, and the land tax (kharaj), imposed on
land, regardless of the landholder’s religion, on the territorial basis (e.g. Abu-Youssef
1979, pp. 122-4). Historians, however, disagreed as to whether this distinction existed in
the early period (640-750).
22

Was the poll tax burden (as a percentage of income) greater at lower income levels?
In 640-750, the annual poll tax was, according to Arabic medieval narratives, a fixed
amount of 2 dinars per person, but the papyri tax documents suggest that this amount was
only an average (Ibn-Abdel-Hakam 1974, pp. 64-6; Morimoto 1981, pp. 66-91).
23
Under

22
The year 750 signifies the accession of the Abbasids to power. Wellhausen (1927, pp. 477-82) and
Becker (1903, pp. 81-112) argued that the tax in 640-750 was, following the Byzantine system, a fixed
tribute imposed at the village-level, with no distinction made between land and poll taxes. The distinction
emerged later on with the fiscal reforms of Umar II in 720. Dennett (1950, pp. 62-103) argued, to the
contrary, that the distinction existed since the Conquest, and that the poll tax was assessed at the individual-
level from the outset. Morimoto (1981, pp. 53-144) suggested that the individual principle was applied
from the beginning to estimate each village’s tribute but only imperfectly and that the poll tax became fully
institutionalized with the fiscal reforms of the first half of the eighth century.
23
Morimoto interprets the individual variation in the poll tax amount in the papyri in 640-750 as reflecting
variation in income. However, it is impossible, based on the information in the tax documents, to confirm
or reject this claim.
150

the Abbasids (750-969), the Fatimids (969-1171), and the Ottomans (1517-1855), the
Sunni Hanafi and the Shiite Ismaili schools of Islamic jurisprudence both dictated that the
poll tax is imposed on adult males according to three income brackets: 1 dinar on manual
workers, 2 dinars on the middle-income, and 4 dinars on the rich (Hanafi: Abu-Youssef
1979, pp. 122-4; Ismaili: Al-Qadi Al-Nu’man 1963, pp. 379-381). Yet, under the
Ayyubids (1171-1250) and the Mamluks (1250-1517), the dominant Sunni Shafi’i School
stressed that the poll tax is fixed (Al-Shafi’i 2001, pp. 423-30), although the authorities
often adhered to the three-brackets system (Mahmoud 2009a; pp. 32-7).
24
Over the whole
period 750-1855, the actual tax assessment was often arbitrary and subject to the
assessor’s discretion although it was generally in line with the theory (Morimoto 1981,
pp. 175-81; Ismail 1998, pp. 164-7; Mahmoud 2009a, pp. 147-81). At any rate, the
limited variation in the poll tax amount under both the three-brackets and the fixed tax
systems meant that the tax burden was larger at lower levels of income. To examine this
argument, I used Ashtor (1969) to obtain occupation-based wages in medieval Egypt, and
I classified each occupation into one of the three poll tax brackets based on the criteria
mentioned in Abu-Youssef (1979, p. 122). Figure 4.2 shows the scatter plot of wages and
poll tax burden under the three-brackets system in 660-1517. The tax burden is quite high
at lower income levels but falls as income increases, thus suggesting that the financial
incentive to convert to Islam was stronger for low-income Coptic taxpayers.

24
The Sunni Hanbali and Maliki Schools do not specify income brackets but they both assert that the poll
tax should be assessed according to the individual’s economic means (Hanbali: Abul-Fadl Saleh 1999, p.
344; Maliki: Malik 1985, pp. 278-80; Al-Baji 1999, pp. 275-9).
151

Were the poor exempt from the poll tax? While the Sunni Hanafi, Hanbali, and
Maliki schools all stated that those living on charity (or are without occupation) are
exempt from the poll tax, both the Shiite Ismaili and the Sunni Shafi’i schools did not
grant such exemption. Noticeably, even under the exemption views, any active adult male
is considered non-poor and is not exempt from the poll tax. Moreover, using evidence
from the Cairo Geniza on destitute Jews who paid the poll tax, Goitein (1963) and
Alshech (2003) argued that the Ayyubids applied the stringent Shafi’i non-exemption
viewpoint.
Were Muslim converts exempt from the poll tax? With the institutionalization of the
poll tax from 750 onwards, Islamic jurists agreed that when a non-Muslim converts to
Islam, he becomes exempt from the poll tax starting from the following fiscal year (e.g.
Abu-Youssef 1979, p. 122). The period 640-750, however, contains examples for
collecting the poll tax even after conversion to Islam (Morimoto 1981, pp. 66-91).
Did the real poll tax amount have a specific trend over the Middle Ages because of
inflation/deflation? I constructed a time series for the purchasing power (in kilograms of
wheat) of the nominal poll tax for the lowest income bracket (1 dinar) over the period
706-1450, using Ashtor (1969, pp. 454-5) (Figure 4.3). The figure shows that the real poll
tax fluctuated over time but that there is no trend in the series.
Having examined the central aspects of poll tax historiography, I turn in subsections
(4.5.2-4.5.5) to the empirical evidence on the origins of the nineteenth century Coptic
spatial distribution. I introduce four historical factors: (i) Tax revolts in the eighth and
ninth centuries and (ii) Arab immigration in the seventh to ninth centuries, as measures of
152

poll tax enforcement, and (iii) Coptic churches and monasteries in the twelfth and
fifteenth centuries and (iv) the route of the Holy Family, as measures of the tax elasticity
of conversion.
4.5.2 Eighth and Ninth Centuries Tax Revolts
Over the period 726-866, a number of anti-government revolts erupted throughout
Egypt, because of higher taxation (poll and land taxes) and stricter tax enforcement
(Morimoto 1981, pp. 145-72; Mikhail 2004, pp. 195-211).
25
Table 4.4 shows the
sequence of these revolts by year, region, reasons, and parties revolting. Early revolts
were led by “disenfranchised Coptic rural elites,” who used to dominate tax collection in
640-750, against the immigrant Arab elites’ increased control of the fiscal administration
(Sijpesteijn 2009). Also, noticeably, the first revolt coincided with the earliest attempts at
institutionalizing poll and land taxes in 720. Later revolts involved both Copts and Arabs,
who both protested against the government’s attempts at centralizing tax collection. The
regional variation of these revolts may thus proxy for poll tax enforcement, which, in
turn, may explain the variation in the Coptic population share in the nineteenth century.
The tax revolts evidence is suggestive, however, since the variation is only at the region-
level. Out of the 15 revolts, 13 took place in the Nile Delta, which has the lowest Coptic
population share (2 percent) in 1897. Only one of the two revolts that occurred in the Nile
Valley, the region with the highest Coptic population share in 1897 (12 percent), was a
pure tax revolt.
26
Perhaps because of its proximity to the capital, and its consequent

25
Sijpesteijn (2009) mentions an early tax revolt that took place in Upper Egypt in 697.
153

receipt of larger Arab immigration waves, the Nile Delta witnessed, on average, stricter
tax enforcement than the Nile Valley (as reflected in the tax revolts), and thus larger
conversions among poor Copts to Islam.
27

4.5.3 Seventh – Ninth Centuries Arab Immigration Waves
I examine the three remaining historical factors that predict the variation in the
Coptic population share in the nineteenth century through estimating the following linear
regression(s):
(2)Coptshare
jk
=α+β
k
+γHistorigin
jk
+μlpop
jk
+ε
Where Coptshare is the percentage of Copts in an area j in greater area k in 1897.
Historigin is the historical origin of interest {seventh to ninth centuries Arab immigration
dummy, Number of Coptic churches and/or monasteries in the twelfth and fifteenth
centuries, Visited by the Holy Family dummy}. β
k
is greater area k fixed effects. An area
j is defined at (i) the district-level for Arab immigration, and at (ii) the urban district- and
rural village- level for the other two origins. A greater area k is defined at the province-
level in (i) and at the urban province- and rural district- level in (ii). The results are
shown in table 4.5. Each historical origin is entered in a separate regression. In subsection
4.5.6, I examine which of these factors was more important. I show the results with and

26
The information on tax revolts does not allow me to break down the Nile Valley into its three regions
(North, Middle, and South).
27
Besides the stricter tax enforcement mechanism, the violent suppression of the tax revolts by the
authorities may have demoralized the Coptic population in the Nile Delta and caused their mass conversion
to Islam. This mechanism was first mentioned by Al-Maqrizi (2002), and is also found in Lane-Poole
(1969), Dennett (1950), and Mikhail (2004).
154

without fixed effects in order to understand the source of variation in the factor of
interest.
Local elites, village headmen, and large landholders played a crucial role in the poll
tax system throughout the institution’s history in Egypt.
28
In the early period (640-750),
Arabs kept the Byzantine system, which relied on the Coptic elites in tax assessment and
collection, mostly intact. Coptic village headmen assessed the individual tax amounts,
which were then aggregated to estimate the village’s tribute (Bagnall 1996, p. 318;
Morimoto 1981, pp. 66-91; Frantz-Murphy 1999). State attempts at centralizing the tax
system started in 720, and strengthened with the accession of the Abbasids to power in
750. In response to the tax revolts, the government resorted to the decentralized tax-
farming system (Sijpesteijn 2009), which remained in effect, in various forms, until the
nineteenth century (Cuno 1992, pp. 17-32). Under this system, the government contracted
out (through auctions) the tax collection of each geographic unit to private individuals,
who then collected the taxes from the individual taxpayers (Morimoto 1981, pp. 231-3).
Under the Ayyubids and the Mamluks (1171-1517), tax-farming took the form of the
Iqta’ system, whereby a military elite was granted large landholdings and control over
local tax collection. However, the village headmen and local elites remained dominant as
the “men on the spot” in the tax collection process whether they operated under a central
government, military elite, or an urban tax-farmer (Cuno 1992, p. 85).

28
Under both the Byzantine and Islamic rule, rural land in Egypt was owned by the state, which gave
usufruct right to private landholders to farm the land in exchange for land-tax, which was technically
considered a “rent.” Only in the late Fatimid period (969-1171), private landownership emerged as rulers
gave away pieces of land to individuals for reclamation.
155

Arab immigration to Egypt was mostly temporary in the seventh century, but
permanent settlement increased in the eighth and ninth centuries. Arab settlers formed a
rural elite with large landholdings, which altered the religious composition of the elites in
the districts they migrated to, and facilitated the “Arabization” of the fiscal administration
(Sijpesteijn 2009). In 833, Arabs lost their position as a military aristocracy to Turks with
the stoppage of their state pensions (Morimoto 1981, p. 167), and Arab immigration to
Egypt dropped after the ninth century as a result. I argue that in the districts that received
Arab immigrants over the seventh to ninth centuries the poll tax was perhaps more
strictly enforced on Copts, leading to larger conversions among poor Copts to Islam. By
contrast, the districts where Arabs did not migrate to and Coptic elites remained in power
may have witnessed lower tax enforcement because of the possible collusion between
Coptic elites and Coptic taxpayers. Arguably, these districts had lower conversions
among poor Copts to Islam.
I examine this possibility in the first row of table 4.5, which shows the impact of the
seventh – ninth centuries Arab immigration wave dummy on the Coptic population share
(at the district-level) in 1897. Consistent with the argument above, I find that the
coefficient is negative and statistically significant, when no province fixed effects are
included. When I control for the latter, however, the coefficient becomes insignificant,
suggesting that the variation in Arab immigration is between, rather than within,
provinces. While this coefficient embodies a direct impact of the Muslim Arab
immigration on a district’s Coptic share, both demographic and historical evidence on the
156

size of Arab immigration confirm that the effect is mostly indirect through the conversion
of Copts to Islam.
4.5.4 Coptic Churches and Monasteries in the Twelfth and Fifteenth Centuries
With the rise of Christianity in Egypt in the 4
th
and 5
th
centuries, Coptic churches and
monasteries played a central role in providing financial and physical aid to the Coptic
population. In the Byzantine period, Bagnall (1996, p. 316) describes how priests in the
4
th
century assumed the new leadership in villages, and Bowman (1989, pp. 117, 129-30)
points out that monasteries provided economic aid and shelter to farmers. After the Arab
Conquest, Coptic churches and monasteries continued to assume the same role. Their
large landholdings were leased out to Coptic farmers for cultivation (Richter 2009).
Papyri evidence preserved loan documents (both with and without interest) provided by
Coptic institutions, or by individual monks and clergymen, to Coptic laymen for the
purpose of tax payment (poll or land tax), as well as in exchange for future services
provided by the debtor to the institution (Markiewicz 2009). Morimoto (1981, p. 118)
provided evidence on Coptic taxpayers who took refuge in the monasteries and pretended
to be monks in order to avoid paying the poll tax.
29
This seems consistent with the
evidence from Cairo Geniza on the support provided by Jewish institutions to the Jewish
community for poll tax payment in the Ayyubid period.
Based on this evidence, I predict that the larger the number of Coptic churches and
monasteries in a geographic unit in the Middle Ages (twelfth and fifteenth centuries), the

29
Monks were exempt from the poll tax, although some rulers in 640-750 taxed them. To control the
monastic population, each monk was branded on his left hand with an iron bracelet identifying his name
and the monastery he belonged to (Morimoto 1981, p. 118).
157

greater is its Coptic population share in 1897. The mechanism is that these geographic
units enjoyed larger support for poor Coptic taxpayers by Coptic institutions and thus
witnessed less conversion among poor Copts since the demand for Islam was less
responsive to the poll tax over there. I expect that the impact of monasteries is likely to be
stronger than churches because the former were richer and possessed large landholdings
in their geographic units. The results are shown in rows 2-7 in table 4.5. Notice that I am
excluding the geographic units that were named after a Coptic monastery in 1897 and that
were thus mostly inhabited by the monastic population. The estimates when I control for
greater area k fixed effects are often different in magnitude from those without fixed
effects indicating that the source of variation could be between or within the greater
areas. I find that there is generally a positive and statistically significant impact of the
number of Coptic churches and monasteries in an area in the twelfth century (and
fifteenth century) on its Coptic population share in 1897, but that the effects in the
fifteenth century are stronger. Also, the effect of monasteries is stronger than that of
churches in the twelfth century but not in the fifteenth century, as predicted by the
historical evidence.
Two endogeneity concerns are important to discuss here. There is a concern about
the “selected” building of new Coptic institutions by the twelfth and fifteenth centuries,
because of survival of Copts in an area. Fortunately, historical evidence suggests that
these Coptic institutions were mostly pre-Islamic, since building new churches and
monasteries was generally restricted under Islamic rule. The other concern here, which is
perhaps more challenging, is the “selected” demolition of Coptic institutions because of
158

Islamization prior to the twelfth and fifteenth centuries. Indeed, one may interpret the
results on churches as suggesting that Islamization took place by the twelfth century and
was further strengthened in the twelfth to fifteenth centuries (Section 4.6.1). However,
the possible demolition of Coptic institutions is mainly a concern for churches but not for
monasteries, which were more likely to survive because of their larger size and wealth. It
is the Coptic monasteries that played the larger role historically in providing aid to poor
Coptic taxpayers.
4.5.5 The Flight of the Holy Family to Egypt
Ancient and medieval Coptic local traditions preserved the route of the biblical flight
of the Holy Family to Egypt and the “miraculous” sites that they touched or created. I
argue that the geographic units that lie on this route may have been more attached to
Christianity, either because of these traditions or as reflected in “inventing” these local
traditions. The poll tax may have thus been less effective in converting poor Copts in
these units, and so they are likely to have higher Coptic population share in 1897. I
examine this possibility in the last row in table 4.5. As expected I find that the coefficient
on the route of the Holy Family dummy is positive and statistically significant.
The historical factors (Coptic tax revolts, Arab immigration waves, Coptic churches
and monasteries, and the route of the Holy Family) thus individually predict the variation
in the Coptic population share in the nineteenth century, which is inversely related to the
size of the Coptic-Muslim gap. I now investigate the interactions between the last three
factors.

159

4.5.6 Which Historical Origins Are More Important?
Table 4.6 shows the estimation results of equation (2), without fixed effects, when
the three historical origins are entered simultaneously, by aggregating the Coptic
churches and monasteries and the route of the Holy Family dummy to the district-level.
All three factors (Arab immigration, Coptic monasteries, and the route of the Holy
Family) retain their expected signs and are statistically significant, although the latter two
factors have smaller coefficients than in table 4.5, presumably because of high
multicollinearity. Coptic churches are significant in the fifteenth century, but not in the
twelfth century (when no fixed effects are included), as in table 4.5.
In table 4.7, I examine the inter-correlation between the historical factors in more
depth. I estimate the determinants of the spatial distribution of Coptic institutions in the
twelfth and fifteenth centuries, in order to understand its relationship with the earlier
historical origins: Arab immigration and the route of the Holy Family. It appears that the
latter factor, as a measure of attachment to Christianity, explains the spatial distribution
of Coptic churches (but not monasteries) in the twelfth and fifteenth centuries.
I now turn to examining the mechanisms through which the impact of the poll tax
institution, which operated over the period 640-1855, persisted until the nineteenth
century.
4.6 Understanding the Mechanisms
4.6.1 When Did the Coptic-Muslim Gap Emerge?
Was the Coptic-Muslim human capital gap born in the nineteenth century? Before
the Conquest, the Coptic majority was, on average, worse off than Melkites (Greeks and
160

Hellenized Egyptians) and Jews. First, Copts were mostly rural, while Jews and Melkites
lived in urban centers. In the twelfth century, only 2 percent of Christian institutions were
Melkite, and 89 percent of them were in the Nile Delta and urban centers.
30
Second, since
Greek was the language of the Byzantine administration, Melkites, who spoke only Greek
(Mikhail 2004, p. 133), had a privilege in white-collar jobs. On the contrary, Copts
worked mostly in farming, although, perhaps notably, there was a bilingual Coptic elite,
that knew both Greek and Coptic, was well-educated, and worked in the bureaucracy.
31

Coptic dominance over administrative occupations before the nineteenth century is
well documented in the historical literature and medieval narratives (e.g. Tagher 1998;
Sheikho 1987; Amer 2000; Samir 1996; Al-Muqaddasi 1877; Al-Maqrizi 2002). Upon
the Conquest, Arabs left the bureaucracy in the hands of the local Christian population
(Butler 1996, p. 465). By the end of the seventh century, Copts, favored by the new Arab
rulers, increasingly replaced Melkites in the bureaucracy, both because of the animosity
between Arabs and Melkites (Byzantines), and the increasing Melkite emigration from
Egypt after the Conquest (Lane-Poole 1969, p. 26; Mikhail 2004, pp. 105-6). Coptic
administrative dominance reached its height in the Fatimid period (969-1171), the so-

30
The statistics are based on my calculations from the dataset on churches and monasteries in the twelfth
century that I constructed from Abul-Makarim (1991). The urban centers are Cairo, Alexandria, and the
deserted city of Tinnis in Northern Delta. This evidence is consistent with Mikhail (2004, p. 134), who
points out that there is no literary or documentary evidence on Melkite presence in the Nile Valley in the
post-Conquest period. Their pre-Conquest situation was not different. In the early seventh century, there
were only 7 Melkite churches in Egypt (Mikhail 2004, p. 48).
31
On the existence of a bilingual Coptic elite in the Byzantine period, it suffices to mention the biographies
of prominent Copts who learned Greek in order to have access to governmental jobs, and the large portion
of the Coptic literature and theology (Coptic here is defined in ethno-religious terms) that was composed in
Greek (Mikhail 2004, pp. 135-48). It was this elite that deliberately created the Coptic script in the third
century to express the Egyptian phonetics with Greek alphabets. See Bagnall (1996, pp. 230-65).
161

called “Golden Age of Non-Muslims,” where Copts, along with Melkites, Jews, and
Armenians, became the highest officials, and not only the bureaucratic machine. In the
Ayyubid and Mamluk period (1171-1517), the attempts to replace Coptic clerks with
Muslims were not successful, and they continued to control administration under the
Ottomans (1517-1914).
32
It suffices here to draw parallels between Al-Muaqqadasi’s
statement, “Scribes in the Levant and Egypt are Christians,” written circa 1000 AD, and
Lord Cromer’s, the British Consul of Egypt (1883-1908), statement, “When the English
took Egyptian affairs in hand, the accountants in the employment of the Egyptian
government were almost exclusively Copts,” that was made a millennium later.
33

Moreover, the papyri provide additional evidence on the persistence of the phenomenon
since the early Middle Ages. Using Maghawri’s (2000) Titles and Names of Crafts and
Jobs in the Light of Arabic Papyri, which traces each occupation in the Arabic papyri, I
find that administrative occupations such as “Gahbaz or Tax Collector,” “Katib or
Scribe,” and “Gistal or Secretary,” are mainly connected with Coptic names.
Having established the persistence of the Coptic administrative monopoly before the
nineteenth century, several questions arise: First, did this monopoly reflect an
“occupational gap” against the growing Egyptian Muslim majority or did Egyptian
Muslims dominate other white-collar jobs that declined, for some reason, before the
nineteenth century? Muslims traditionally dominated four “white-collar” occupations: the

32
Tagher (1998, p. 142) described the “persistence” of the phenomenon as follows: “The condition of the
Copt did not change during the six centuries preceding (the nineteenth century)… His work, tax collecting,
was the basis of his existence and his only hope to accumulate wealth” (Italics mine).
33
The statements are from Al-Muqqadasi (1877, p. 183) and Tagher (1998, p. 213).
162

military, the judiciary, the police, and the ulama (Muslim clergy). As the nineteenth
century is known in Egyptian history as a period of unprecedented military expansion and
modernization, there is no reason to believe that the relative size of these Muslim white-
collar occupations was smaller than before. Moreover, starting from Al-Mu’tasim’s reign
(833-42), officers’ ranks in the regular army became controlled by Turks, i.e. non-
Egyptian Muslims (Morimoto 1981, p. 160). The hostile attitude that Muslims expressed
towards the dominance of non-Muslims over the administration in the Middle Ages
perhaps indicates that this dominance was not compensated for by other Muslim white-
collar venues.
34
In addition, evidence from Ashtor (1969) suggests that wages of
governmental officials exceeded, on average, those of artisans and laborers, and thus this
occupational gap reflected an income gap.
Second, when did this occupational gap emerge? The timing of the emergence of the
gap is, to an extent, related to the timing of Copts becoming a minority (i.e. less than 50
percent of the population). My finding that the spatial distribution of Coptic churches in
the twelfth and fifteenth centuries predicts the nineteenth century Coptic spatial
distribution provides suggestive evidence on this point. It may be safe to claim that the
nineteenth century geographic pattern of Islamization was determined by the twelfth

34
Examples for this hostile attitude include the verse by the tenth century Muslim poet, Al-Hassan Ibn-
Bishr Al-Dimashqi, in which he mocks three top Christian officials in Egypt at the time, “Convert to
Christianity for Christianity is the true religion! This is evidenced by our era! Believe in three who
achieved lordliness and glory! Do not believe in others who are idle! The Father is Jacob the minister, Aziz
is the Son, and the Holy Spirit is Fadl,” quoted in Sheikho (1987, p. 20). Other examples in Tagher (1998,
p. 36) include the medieval books that sought a “theoretical” basis in Islamic jurisprudence for prohibiting
non-Muslims from working in the Muslim government. These include the fourteenth century books by Ibn-
Al-Naqqash, Reprehensibility of Hiring Ahl-Al-Dhimma (Christians and Jews), and Ibn-Al-Durayhim’s
The Right Way to Explain the Ugliness of Hiring the People of the Book (Christians and Jews).
163

century, but was strengthened between the twelfth and the fifteenth centuries. Historical
evidence suggests that Coptic regional distribution (their relative concentration in the
Nile Valley rather than the Nile Delta) predates the twelfth century. A liturgical Coptic
text from the eighth to tenth centuries mentioned that, “the remaining (Copts) in Upper
Egypt (Nile Valley) who know the Coptic language and speak it are mocked and
insulted.”
35
This evidence along with my findings lend support to the two existing
historical viewpoints on Islamization of Egypt: Islamization took place before the twelfth
century but received a “push” in the Mamluk period (1250-1517). Thus, the Coptic-
Muslim gap likely emerged before the twelfth century, and widened over time with
continued selected conversion to Islam.
Finally, what occurred to Jews and Melkites, who were, in the pre-Conquest period,
better off than Copts? Conceptually, these groups were perhaps less affected by the poll
tax not only because of their ex ante better economic position but also because of their
smaller size that facilitated solidarity in poll tax payment as evidenced by the Cairo
Geniza (Alshech 2003). Interestingly, in 1848, there seem to be no changes in the size or
in the relative economic status of these groups: They remained two small minorities that
are better off than both Copts and Muslims.
36

35
This text is the apocalypse of Samuel, Bishop of the Monastery of Qalamun (Ziadeh 1915-17). The
document is undated, but it is suggested that it dates from the eighth to tenth centuries (Papaconstantinou
2007).
36
While it is correct to assume that nineteenth century Copts are the descendants of the pre-Conquest
Coptic population, since Copts were an isolated and highly immobile group throughout the whole period, it
is more difficult to make this claim for Jews and Melkites (Orthodox Ruum). The two latter groups in
Egypt received large immigration waves from Spain and other parts of the Islamic world.
164

4.6.2 Why Did the Gap Persist?
Medieval institutions in Egypt were gap-preserving as they limited occupational and
geographic mobility. The hereditary transmission of occupations was the norm in the pre-
modern period. Tagher (1998, pp. 212-3) and Samir (1996) described the barriers to entry
set up by Coptic scribes in order to maintain their monopoly of administrative jobs. They
insisted on using the more difficult fraction system (rather than decimals) in accounting.
In the words of Lord Cromer, the Coptic accounting system was “archaic,” and they
resisted all attempts at reform or simplification of the system.
37
Its “secrets” were
transferred from one generation to another, as Coptic children accompanied their fathers
to governmental offices to learn.
Geographic mobility was also constrained and led to the persistence of the spatial
distribution of Copts and the Coptic-Muslim gap over the centuries. The central
government in Egypt controlled the movement of individuals since antiquity. Morimoto
(1981) mentions the problem of fugitives, who fled their villages to avoid paying the poll
and land taxes in the seventh – tenth centuries, and the state’s efforts to control their
movement. Mahmoud (2009a, pp .159-60) and Cuno (1992, pp. 121-4) make similar
observations for Ottoman and nineteenth century Egypt.
4.6.3 Other Historical Explanations for the Coptic-Muslim Gap
In this subsection, I examine other possible explanations of the conversion of Copts
to Islam and the emergence of the Coptic-Muslim economic gap. These include forced

37
The system was also corrupt. Copts manipulated the measurement rod (qasaba) used in calculating the
acreage for land-tax assessment.
165

conversion, land tax, access to governmental jobs, Islamic alms, and Islamic inheritance
laws. In what follows I discuss briefly each of these points.
One possible explanation of the phenomenon is that forced conversion, as a result of
state persecution, may have led to the conversion of Copts to Islam. Also, if poor Copts
were more likely to suffer from persecution and thus to convert, it may also explain the
Coptic-Muslim gap. Historical evidence suggests, however, that forced conversion cannot
explain Islamization in the period 640-1250, where it occurred in two limited episodes
under Al-Muttawakil (847-61) and Al-Hakim (996-1021), and that it only became
common in the Mamluk period (1250-1517). Since the evidence in this chapter suggests
that Islamization and the Coptic-Muslim gap emerged by the twelfth century, it is perhaps
safe to rule out forced conversion in explaining the gap. At most, forced conversion in the
twelfth to fifteenth centuries may have given the phenomenon a further “push.”
Another explanation is that the land tax (kharaj) may have acted as another financial
incentive for Copts to convert to Islam. However, in Islamic jurisprudence, land tax is not
different across Muslims and non-Muslims, for it is assessed on the territorial basis
regardless of the religion of the landholder. In the early period (640-750) though, the land
tax was lower for Muslims, and so may have provided an additional incentive to convert.
One may also argue that the conversion of Copts to Islam occurred because of the
desire to access high governmental jobs, which were preserved for Muslims. However,
this incentive could have operated only among the Coptic elite, as is attested to in the
historical literature, and thus does not explain the positive Coptic-Muslim gap.
166

One counterargument against the poll tax hypothesis is that Muslims were subject to
other financial obligations such as the alms (zakat), which may have counterbalanced the
poll tax incentive to convert. However, poor Muslims were not only exempt from the
alms, but also recipients of them. Thus, the alms may have been an additional financial
incentive for a poor Copt to convert to Islam.
Finally, one may argue that large Coptic landholders may have preferred to remain
Christian in order to avoid the Islamic inheritance laws that would fragment their
landholdings among their heirs. However, historical and quantitative evidence suggest
that the bulk of the Coptic-Muslim gap is not because of the existence of a large Coptic
landholding elite, that would be affected by the Islamic inheritance system, but because
of the large Coptic bureaucracy who earned their living from governmental wages.
4.7 Conclusion
Military conquests and exporting the victorious side’s institutions to the conquered
territories may have lasting impact on the defeated populations, even if this impact was
not part of the original intention of the exported institution. Islamic Conquest of Egypt in
the seventh century AD installed a new fiscal system in the country, and the Islamic poll
tax institution was a major part of this system. The tax was imposed on non-Muslims who
were exempt from the tax once they converted to Islam. Unintentionally, the tax led to
the widespread conversion of poor Copts to Islam to avoid paying the tax, and thus to the
shrinking of Copts into a well-educated and better off minority than the growing Muslim
majority. In this chapter, I provided quantitative evidence that the historical spatial
variation in the enforcement of the poll tax and the tax elasticity of conversion, as
167

reflected in four historical factors, explains the nineteenth century variation in the Coptic
population share. This is, in turn, inversely related to the magnitude of the Coptic-Muslim
gap. The four historical origins that I focused on are: the tax revolts in the eighth and
ninth centuries, the Arab immigration waves to Egypt in the seventh to ninth centuries,
the Coptic churches and monasteries in the twelfth and fifteenth centuries, and the route
of the Holy Family. I drew on a wide range of qualitative evidence to support and
interpret my findings.
By providing a historical case study where the currently observed correlation
between religion and economic outcomes is due to the historical self-selection into the
religion, rather than religious affiliation per se, the chapter contributes to the current
economics of religion literature (e.g. Botticinni and Eckstein 2002 and 2005). Also, the
chapter contributes to the ongoing historical literature on the Islamization of Egypt and
the Middle East (e.g. Mikhail 2004; Al-Leithy 2005). The new data sources that I
digitized on Coptic churches allowed me to present suggestive evidence that Islamization
of Egypt took place before the twelfth century and that the Mamluk period (1250-1517),
with the state’s pressure on non-Muslims, gave Islamization a further and strong push.
The chapter raises new, and perhaps more challenging, questions: Did the Islamic
poll tax institution that operated in other countries in the Middle East, and the European
parts of the Ottoman Empire, produce similar religious human capital gaps? Can the
religious human capital gaps that are observed in Lebanon and Syria for example be
explained similarly? Why did Christianity survive in Egypt, Lebanon, Syria, and Iraq, but
was completely wiped out from North Africa? Can this be explained by cross-country
168

variation in the Islamic poll tax institution or is it due to other factors? There are new data
sources that may shed light on at least some of these questions: The tax registers (Tahrir
Defterleri) from sixteenth century Ottoman Syria, Palestine, and Transjordan (Hütteroth
and Abdulfattah 1977) include information on poll tax amount and religious composition
at the village-level. The recently discovered 1528 cadastral survey from Ottoman Egypt
described in Mahmoud (2009b) may include similar information. When combined with
the digitized samples of the 1848/68 Egyptian censuses, and the non-digitized Ottoman
individual-level censuses from Lebanon, Syria, and Palestine in 1881-92 (Karpat 1978),
these sources can perhaps deepen our understanding of the historical process of
Islamization of the Middle East.

169

Figure 4.1: Social Status Index Distribution by Religious Group in
Rural and Urban Egypt (1848 and 1868 Pooled Individual-Level
Digitized Samples)
0%
10%
20%
30%
40%
50%
60%
70%
A. Rural Egypt
Muslims
Copts

0%
10%
20%
30%
40%
50%
60%
B. Urban Egypt
Muslims
Copts

The figures are based on the pooled individual-level digitized samples from the 1848 and 1868 Egyptian
censuses. See Chapter 2 and Appendix C for the construction of SSI and the definition of each value.
170

Figure 4.2: Wages and Poll Tax Burden in Medieval Egypt

The figure is based on Ashtor (1969). Each point represents an occupational title. The annual wage of an
occupational title in each time period is the average annual wage across the observations mentioned in
Ashtor (1969) practicing the occupation in the specified time period. Each occupational title is classified
into one of the three poll tax brackets according to Abu-Youssef (1979, pp. 122-4), in order to assign the
annual poll tax amount. The poll tax burden is defined as the quotient of the annual poll tax amount over
the annual wage in current dinars.

171

Figure 4.3: Purchasing Power (in Kilograms of Wheat) of the
Lowest Bracket Annual Poll Tax, One Dinar (706-1450)

0
50
100
150
200
250
706
800
876
889
920
968
1000
1024
1050
1100
1157
1200
1201
1220
1232
1239
1250
1295
1296
1300
1325
1349
1400
1450
Purchasing Power in Kilograms of Wheat
Year

The figure is based on the information on prices in Ashtor (1969).

172

Table 4.1: Descriptive Statistics: Individual-Level 1848 and 1868
Census Samples and 1897 Village/Urban Quarter-Level Full
Census

1848/68 1897
Social Status Index: Mean (Standard Deviation) (Min = 1, Max = 12)
Urban Provinces
Muslims (N = 22,700) 5.40 (2.40) -
Copts (N = 827) 7.69 (1.79) -
Rural Provinces
Muslims (N = 11,642) 5.05 (1.72) -
Copts (N = 916) 5.83 (1.85) -
National Religious Distribution
% Muslims 91% 92%
% Copts 6% 6%
% Other Non-Muslims 1% 2%
% Unspecified Religion 2% 0%
% Copts (Out of the Region's Population) by Region
Urban Provinces 3% 3%
Nile Delta 2% 2%
Northern Nile Valley 4% 5%
Middle Nile Valley 18% 19%
Southern Nile Valley 3% 6%
Regional Distribution of Copts
Urban Provinces 9% 6%
Nile Delta 13% 15%
Northern Nile Valley 10% 8%
Middle Nile Valley 65% 62%
Southern Nile Valley 3% 10%

Statistics in the first column are computed from the 1848 and 1868 pooled individual-level census
sample which is restricted to adult active males who are at least 15 years old in 1848/68. The
oversample of non-Muslims in Cairo is not used in the descriptive statistics but is used in the
regressions in table 4.3. Sampling weights are used to account for the different sampling rates
across provinces. Statistics in the second column are computed from the 1897 village-level full
census. See Chapter 2 and Appendix C for details on the non-Muslims' oversample and the
construction of the Social Status Index, Chapter 1 for details on the sampling methodology, and
Chapter 3 for details on the 1897 census data.

173

Table 4.2: Descriptive Statistics: Medieval Arab Immigration
Waves and Churches and Monasteries of Egypt

1. 7th - 9th Centuries Arab Immigration Waves
Number of Arab immigration waves 296
% in unknown destination (excluded from sample) 1%
% in temporary non-settlement migration (Irtiba') (excluded from sample) 17%
Origin of Arab Migration Waves in the Sample (N = 244)
% Adnan tribes immigration waves (from North Arab Peninsula) 31%
% Qahtan tribes immigration waves (from South Arab Peninsula) 66%
Destination of Arab Migration Waves in the Sample (N = 244)
% in Urban Provinces 49%
% in Lower Egypt 25%
% in Upper Egypt 26%
% 1897-Districts that received at least one Arab immigration wave in the 7th- 9th
centuries
42%
2. 12th Century Churches and Monasteries
Number of churches 1,118
Number of monasteries 136
% in deserted or unknown village (excluded from sample) 9%
% Recently Demolished by the 12th century (excluded from sample) 3%
Denomination of Institutions in the Sample (N = 1,110)
% Coptic 96%
Spatial Distribution of Coptic Churches and Monasteries in the Sample (N = 1,062)
% in Urban Provinces 13%
% in Lower Egypt 42%
% in Upper Egypt 45%
% 1897-Villages that existed before 1477, are not Coptic monasteries in 1897, and
had at least one Coptic church/monastery in the 12th century
15%

174

Table 4.2 (Continued)

3. 15th Century Churches and Monasteries
Number of churches 201
Number of monasteries 87
% in deserted or unknown village (excluded from sample) 11%
% Recently Demolished by the 15th century (excluded from sample) 43%
Denomination of Institutions in the Sample (N = 154)
% Coptic 93%
Spatial Distribution of Coptic Churches and Monasteries in the Sample (N = 145)
% in Urban Provinces 21%
% in Lower Egypt 10%
% in Upper Egypt 69%
% 1897-Villages that existed before 1477, are not Coptic monasteries in 1897, and
had at least one Coptic church/monastery in the 15th century
3%
4. Route of the Holy Family
Number of sites visited by the Holy Family 38
% in deserted or unknown village (excluded from sample) 3%
Spatial Distribution of the Sites in the Sample (N = 37)
% in Urban Provinces 11%
% in Lower Egypt 32%
% in Upper Egypt 57%
% 1897-Villages that existed before 1477, are not Coptic monasteries in 1897, and
are believed to have been visited by the Holy Family
1%
Source: Digitized samples are described in section 4.3 in the chapter.

175

Table 4.3: Coptic Population Share and Coptic-Muslim
Economic Gap in 19th Century Egypt
Linear Individual-Level Regression-
Dependent Variable: Social Status Index-
Pooled Individual-Level 1848 and 1868 Census Samples
(1) Area =
Urban District
and Rural
Village
(2) Area
= District
(3) Area =
Province
(4) Area =
Region
Coptic Christian 2.030***
(0.109)
2.143***
(0.164)
2.363***
(0.161)
2.437***
(0.168)
% Copts in the Area (1897) 0.555
(0.383)
0.564
(0.910)
6.580
(6.143)
0.009
(0.533)
Coptic Christian × % Copts
in the Area (1897)
-2.661***
(0.618)
-3.693***
(1.249)
-9.172***
(1.225)
-11.141***
(1.332)
Log (population) (1897) -0.215**
(0.087)
-0.334**
(0.152)
0.180
(0.136)
0.170**
(0.074)
# Individual Observations 36,570 33,747 37,462 37,462
Unit of Geography Urban District +
Rural Village
District Province Region
# Geographic Units 871 68 19 5
Fixed Effects Urban Province
+ Rural District
Province Region No
Adjusted R-Squared 0.073 0.074 0.068 0.062

Each column represents a separate linear regression where the percentage of Copts is measured
at a different geographic level in each. Column 1 represents the smallest geographic level and
column 4 represents the largest. Robust standard errors clustered at the urban district/village-
level are in parentheses. Social Status Index takes the values 1 through 12. Sample is restricted to
Muslim and Coptic adult active males who are at least 15 years old in 1848/68. Numbers of
observations, villages, and urban and rural districts vary across regressions because of the
varying quality of matching between the 1848/68 units and the 1897 units. * indicates
significance at the 10 percent level, ** indicates significance at the 5 percent level, and ***
indicates significance at the 1 percent level.

176

Table 4.4: Coptic Regional Variation in Nineteenth Century Egypt
and Tax Revolts in the Eighth and Ninth Centuries

Year Location Reason Parties Revolting
726 Nile Delta Tightening state control over the tax
system
Copts
740 Nile Valley Higher tax enforcement, collecting
poll tax from fugitives, higher tax
rate, uniform regardless of wealth
Copts
750 Nile Delta Heavy taxation and general suffering Copts, Arabs also
revolted to overthrow
the Umayyads
753 Nile Delta Reorganizing the tax system under the
Abbasids and heavy taxation
Copts
768 Nile Delta Al-Mansour's fiscal institutions Copts
783 Nile Valley Political revolt, proclaiming
leadership over Upper Egypt, in
essence tax resistance
Arabs (Umayyad
descendants), Copts
joined the troops
784 Nile Delta Increase in taxes to be equal to that of
Copts and corruption in collection
Arabs
794 Nile Delta Increase in taxes Arabs
802 Nile Delta Strict tax enforcement Arabs
806 Nile Delta Strict tax enforcement Arabs
807 Nile Delta Strict tax enforcement, heavy taxes Copts led by Arab
landlords
818 Nile Delta High taxes (part of the civil war under
Al-Amin and Al-Ma'moon)
Copts and Arabs
829 Nile Delta High taxes and strict tax enforcement
(part of the civil war under Al-Amin
and Al-Ma'moon)
Arabs then both Arabs
and Copts
866 Nile Delta Taxes Arabs and Copts
866 Nile Delta High Taxes and fiscal reforms by Ibn-
Al-Muddabir
Arabs and Copts against
the new Turkish elites

References: Morimoto (1981, pp. 145-72) and Mikhail (2004, pp. 195-211)

177

Table 4.5: Historical Origins of Coptic Spatial Distribution in
Nineteenth Century Egypt
(Linear Regressions: Dependent Variable: Percentage of Copts in 1897 Census)

Historical Factor (Regressor
of Interest)
Without
Fixed
Effects
With
Fixed
Effects
Level of
Analysis
Level of Fixed
Effects
N
(1) 7th - 9th Centuries Arab
Immigration Wave Dummy
-0.030***
(0.013)
(0.144)
0.005
(0.012)
(0.593)
District Province 126
(2) Number of Coptic
Churches and Monasteries
in the 12th Century (Total)
0.001
(0.001)
(0.027)
0.001*
(0.001)
(0.425)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113
(3) Number of Coptic
Churches in the 12th
Century
0.001
(0.001)
(0.026)
0.002*
(0.001)
(0.425)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113
(4) Number of Coptic
Monasteries in the 12th
Century
0.013**
(0.005)
(0.029)
0.005
(0.004)
(0.424)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113
(5) Number of Coptic
Churches and Monasteries
in the 15th Century (Total)
0.034***
(0.004)
(0.066)
0.024***
(0.003)
(0.442)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113
(6) Number of Coptic
Churches in the 15th
Century
0.045***
(0.004)
(0.072)
0.031***
(0.004)
(0.444)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113
(7) Number of Coptic
Monasteries in the 15th
Century
0.023***
(0.008)
(0.030)
0.019***
(0.007)
(0.426)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113
(8) Visited by the Holy
Family Dummy
0.060***
(0.017)
(0.031)
0.020
(0.014)
(0.425)
Urban District +
Rural Village
Urban Province
+ Rural District
2,113

(i) Coefficient (Standard Error) (Adjusted R-Squared) are reported in the second and third columns.
(ii) Each cell comes from a separate linear regression. All regressions control for the logarithm of
population in 1897 at the specified level of analysis. In rows 2-8, rural villages are restricted to the pre-
1477 villages according to Ramzi (1994) and Ibn-Al-Gay'an (1898) in order to have a map of rural
Egypt that resembles rural Egypt's map in the Middle Ages. Also, in rows 2-8, villages formed mostly
of Coptic monasteries (defined as villages named after Coptic monasteries) are excluded. These two
restrictions resulted in dropping 43 percent of the 1897-villages. * indicates significance at the 10
percent level, ** indicates significance at the 5 percent level, and *** indicates significance at the 1
percent level.

178

Table 4.6: Which Historical Origins Are More
Important?
District-Level Linear Regression- Dependent Variable: Coptic
Population Share in 1897

(1) (2) (3) (4)
Arab Immigration (640-900) -0.038***
(0.013)
-0.040***
(0.013)
-0.031**
(0.012)
-0.036***
(0.013)
Coptic Churches (1200) -0.000
(0.001)
- - -
Coptic Monasteries (1200) - 0.007**
(0.003)
- -
Coptic Churches (1500) - - 0.013***
(0.003)
-
Coptic Monasteries (1500) - - - 0.011**
(0.005)
Visited by the Holy Family 0.041**
(0.016)
0.035**
(0.015)
0.016
(0.015)
0.036**
(0.015)
Adjusted R
2

0.176 0.210 0.298 0.265
N
126 126 126 126

I control for log (population) in 1897 in each regression. * indicates significance at
the 10 percent level, ** indicates significance at the 5 percent level, and *** indicates
significance at the 1 percent level.
179

Table 4.7: Determinants of the Spatial Distribution of Medieval
Coptic Institutions

District-Level Linear Regression- Dependent Variable Indicated on Top of
Each Column

(1) Coptic
Churches
(1200)
(2) Coptic
Monasteries
(1200)
(3) Coptic
Churches
(1500)
(4) Coptic
Monasteries
(1500)
Arab Immigration (640-
900)
0.091
(2.211)
0.220
(0.427)
-0.607
(0.407)
-0.190
(0.240)
Visited by the Holy Family 6.226**
(2.577)
0.400
(0.497)
1.718***
(0.475)
0.192
(0.280)
Adjusted R
2

0.167 0.054 0.129 -0.007
N
126 126 126 126

I control for log (population) in 1897 in each regression. * indicates significance at the 10
percent level, ** indicates significance at the 5 percent level, and *** indicates significance at
the 1 percent level.

180

Bibliography

Abrahams, I. (1896), Jewish Life in the Middle Ages, New York: Macmillan Company.
Abul-Fadl Saleh (1999), Masa’il al-imam Ahmed Ibn-Hanbal (Arguments of Al-Imam
Ahmed Ibn-Hanbal), Al-Riyadh: Dar Al-Watan.
Abul-Makarim (1999), Churches and Monasteries of Egypt, Two Volumes, Anba-Samuel
(Ed.), Cairo: Dar-Al-Na’am.
Abu-Youssef (1979), Kitab al-kharaj (The Book of Land Tax), Beirut: Dar Al-Ma’rifa.
Acemoglu, D., D. Cantoni, S. Johnson, and J. Robinson (forthcoming), “The
Consequences of Radical Reform: The French Revolution,” American Economic Review.
Acemoglu, D., S. Johnson, and J. Robinson (2001), “The Colonial Origins of Economic
Development: An Empirical Investigation,” American Economic Review, Vol. 91, pp.
1369-1401.
Acemoglu, D. (2002), “Technical Change, Inequality, and the Labor Market,” Journal of
Economic Literature, Vol. 40 (1), pp. 7-72.
Al-Baji (1999), Al-muntaqa: sharh muwatta’ Malik (The Chosen: Interpreting Malik’s
Muwatta’), M. Atta (Ed.), Volumes 3 and 4, Beirut: Dar Al-Kutub Al-‘Ilmiya.
Al-Barry, A. (1992), Al-qaba’il al-‘arabiya fi misr fil qoroun al-thalatha al-oula lil hijra
(Arab Tribes in Egypt in the First Three Hijri Centuries), Cairo: Egyptian General Book
Authority.
Alesina, A. and E. La Ferrara (2005), “Ethnic Diversity and Economic Performance,”
Journal of Economic Literature, Vol. 43(3), pp. 762-800.
Al-Gritli, A. (1952), Tarikh al-sina’a fi Misr fil nisf al-awwal min al-qarn al tasi’ ‘ashar
(History of Industry in Egypt in the First Half of the Nineteenth Century), Dar El-
Ma’aref, Cairo.
Al-Hitta, A. A. (1967), Tarikh Misr al-iqtisadi fil qarn al tasi’ ‘ashar (Economic History
of Egypt in the Nineteenth Century), Al-Masry Publishing, Cairo.
Alleaume, G. and P. Fargues (1998), “La Naissance d'une Statistique d'Etat,” Histoire &
Mesure, Vol. 13, pp. 147-193.
181

Al-Leithy, T. (2005), Coptic Culture and Conversion in Medieval Cairo 1293-1524 AD,
PhD Dissertation in History, Princeton University.
Al-Maqrizi (2002), Al-mawa’iz wal i’tibar fi zhikr al-khitat wal athar (Lessons in
Examining Plans and Monuments), Four Volumes, A. F. Sayyid (Ed.), London: Al-
Furqan Islamic Heritage Foundation.
Al-Muqqadasi (1877), Ahsan al-taqaseem fi ma’rifat al-aqaleem (The Best of
Compositions in the Knowledge of Regions), Leiden: Brill Publishing.
Al-Qadi Al-Nu’man (1963), Da’aim Al-Islam (Foundations of Islam), Vol. 1, A. Faydy
(Ed.), Cairo: Dar Al-Ma’arif.
Al-Raf’i, A. (1987), ‘Asr Ismail (The Era of Ismail), Vol. 1, Dar El-Ma’aref, Cairo.
Al-Shafi’i (2001), Al-umm (The Comprehensive), Vol. 5, R. Abdel-Muttalib (Ed.), al-
Mansoura: Dar Al-Wafa’.
Alshech, E. (2003), “Islamic Law, Practice, and Legal Doctrine: Exempting the Poor
from the Jizya under the Ayyubids (1171-1250),” Islamic Law and Society, Vol. 10(3),
pp. 348-375.
Amer, F. (2000), Tarikh ahl-al-dhimma fi misr al-islamiyya min al-fath al-araby ila
nihayat al-‘asr al-fatemy (History of Christians and Jews in Islamic Egypt from the Arab
Conquest to the End of the Fatimid Era), Two volumes, Cairo: Egyptian General Book
Authority.
Anba-Bishoy, P. (1999), The Flight of the Holy Family to Egypt, Cairo: Nubar Printing
House.
Ashtor, E. (1969), Histoire des Prix et des Salaires dans l’Orient Médiéval, Paris: Ecole
Pratique des Hautes Etudes.
Atack, J., F. Bateman, and R. Margo (2004), “Skill Intensity and Rising Wage Dispersion
in Nineteenth-Century American Manufacturing,” Journal of Economic History, Vol. 64
(1), pp. 172-192.
Atiya, A. S. (2005), Tarikh al-massihiya al-sharqiya (History of Eastern Christianity),
Ishaq Ebeid (Tr.), Cairo: Supreme Council of Culture.
Baer, G. (1964), Egyptian Guilds in Modern Times, Israel Oriental Society, Jerusalem.
Bagnall, R. (1996), Egypt in Late Antiquity, Princeton: Princeton University Press.
182

Barkan, Ö. (1957), “Essai sur les Données Statistiques des Registres de Recensement
dans l'Empire Ottoman aux XVe et XVIe Siècles,” Journal of the Economic and Social
History of the Orient, Vol. 1, pp. 9-36.
Barro, R. J. and R. M. McCleary (2003), “Religion and Economic Growth across
Countries,” American Sociological Review, Vol. 68(5), pp. 760-781.
________ (2005), “Which Countries Have State Religions?” Quarterly Journal of
Economics, Vol. 120(4), pp. 1331-1370.
Becker, C. (1903), Beitraege zur Geschichte Aegyptens unter dem Islam, Vol. 2,
Strassburg: Verlag Von Karl J. Trübner.
Becker, S. O. and L. Woessmann (2007), “Was Weber Wrong? A Human Capital Theory
of Protestant Economic History,” CESifo Working Paper No. 1987 and IZA Discussion
Paper No. 2886.
Beinin, J. and Z. Lockman (1987), Workers on the Nile: Nationalism, Communism,
Islam, and the Egyptian Working Class, 1882-1954, Princeton University Press,
Princeton, NJ.
Black, S. E. and L. M. Lynch (1996), “Human-Capital Investments and Productivity,”
American Economic Review, Vol. 86 (2), pp. 263-7.
Boktor, A. (1936), School and Society in the Valley of the Nile, Elias' Modern Press,
Cairo.
____________ (1963), The Development and Expansion of Education in the United Arab
Republic, The American University in Cairo Press, Cairo.
Boppart, T., J. Falkinger, V. Grossmann, and U. Woitek (2008), “Qualifying Religion:
The Role of Plural Identities for Educational Production,” Institute for Empirical
Research in Economics University of Zurich Working Paper No. 360 and CESifo
Working Paper No. 2283.
Borooah, V. K. and S. Iyer (2005), “Vidya, Veda, and Varna: The Influence of Religion
and Caste on Education in Rural India,” Journal of Development Studies, Vol. 41(8), pp.
1369- 1404.
Botticini, M. and Z. Eckstein (2002), “From Farmers to Merchants: A Human Capital
Interpretation of Jewish Economic History,” IZA Discussion Paper No. 670.
________ (2005), “Jewish Occupational Selection: Education, Restrictions, or
Minorities?” Journal of Economic History, Vol. 65(4), pp. 922-48.
183

Bowman, A. (1989), Egypt after the Pharaohs: 332 BC- 642 AD From Alexander to the
Arab Conquest, Berkeley: California University Press.
Bulliet, R. (1979), Conversion to Islam in the Medieval Period: An Essay in Quantitative
History, Cambridge: Harvard University Press.
Butler, A. G. (1996), Fath al-arab li misr (Arab Conquest of Egypt), M. Abu-Hadid (Tr.),
Cairo: Madbuli Library.
Centre d’Etudes et de Documentation Economiques, Juridiques, et Sociales (CEDEJ)
(2003), “Century Census CD-ROM: Egypt 1882- 1996.”
Charmes, G. (1883), Five Months at Cairo and in Lower Egypt, William Conn
(Translator), Richard Bentley & Son, London.
Chaudhary, L. and J. Rubin (2009), “Reading, Writing, and Religion: Explaining Hindu
and Muslim Literacy Differences in Colonial India,” Unpublished Manuscript.
Cochran, J. (1986), Education in Egypt, Croom Helm Ltd., Beckenham, Kent.
Cosgel, M. M. (2004), Ottoman Tax Registers (Tahrir Defterleri), Historical Methods,
Vol. 37, pp. 87-100.
Costa, D. and M. Kahn (2006), “Forging a New Identity: The Costs and Benefits of
Diversity in Civil War Combat Units for Black Slaves and Freemen,” Journal of
Economic History, Vol. 66(4), pp. 936-62.
Courbage, Y. and P. Fargues (1997), Christians and Jews under Islam, J. Mabro (Tr.),
London- New York: I. B. Tauris Publishers.
Cuno, K. (1992), The Pasha’s Peasants: Land, Society, and Economy in Lower Egypt
1740-1858, Cambridge: Cambridge University Press.
Cuno, K. M. and M. J. Reimer (1997), The Census Registers of Nineteenth-Century
Egypt: A New Source for Social Historians, British Journal of Middle Eastern Studies,
Vol. 24, pp. 193-216.
Cutler, D. and E. Glaeser (1997), "Are Ghettos Good or Bad?" Quarterly Journal of
Economics, Vol. 112(3), pp. 827-72.
Dennett, D. (1950), Conversion and the Poll Tax in Early Islam, Cambridge, MA.
Duflo, E. (2001), “Schooling and Labor Market Consequences of School Construction in
Indonesia: Evidence from an Unusual Policy Experiment,” American Economic Review,
Vol. 91 (4).
184

Durkheim, E. (1984), The Division of Labor in Society, Translated by: W. D. Halls, Free
Press, New York.
Fahmy, K. (1998), All the Pasha’s Men: Mehmed Ali, his Army, and the Making of
Modern Egypt, Cambridge University Press, Cambridge.
_______ (2009), Mehmed Ali: From Ottoman Governor to Ruler of Egypt, Oxford: One
World.
Fahmy, M. (1954), La Revolution de l’Industrie en Egypte et Ses Consequences Sociales
au 19e Siècle (1800-1850), E. J. Brill, Leiden, Netherlands.
Fargues, P. (2003), Family and Household in Mid-Nineteenth-Century Cairo, In Family
History in the Middle East, edited by B. Doumani, New York: State University of New
York Press.
Frantz-Murphy, G. (1999), “Land Tenure in Egypt in the First Five Centuries of Islamic
Rule (Seventh-Twelfth Centuries AD),” in Agriculture in Egypt from Pharaonic to
Modern Times, A. Bowman and E. Rogan (Eds.), Oxford: Oxford University Press, pp.
237-66.
Gabra. G. (2001), Be Thou There: The Holy Family’s Journey in Egypt, Cairo: American
University in Cairo Press.
Gerschenkron, A. (1962), Economic Backwardness in Historical Perspective: A Book of
Essays, Harvard University Press, Cambridge, MA.
Ghazaleh, P. (1999), Masters of the Trade: Crafts and Craftspeople in Cairo, 1750-1850,
The American University in Cairo Press, Cairo, Egypt.
Goitein, S. D. (1963), “Evidence on the Muslim Poll Tax from Non-Muslim Sources: A
Geniza Study,” Journal of the Economic and Social History of the Orient, Vol. 6(3), pp.
278-95.
Goldin, C. and K. Sokoloff (1982), “Women, Children, and Industrialization in the Early
Republic: Evidence from the Manufacturing Censuses,” Journal of Economic History,
Vol. 42 (4), pp. 741-774.
Goldin, C. and L. F. Katz (1995), “The Decline of Non-Competing Groups: Changes in
the Premium to Education, 1890 to 1940,” NBER Working Paper SEeries, Working
Paper No. 5202.
_______ (1998), “The Origins of Technology-Skill Complementarity,” Quarterly Journal
of Economics, Vol. 113 (3), pp. 693-732.
185

Guiso, L., P. Sapienzad and L. Zingales (2003), “People’s Opium? Religion and
Economic Attitudes,” Journal of Monetary Economics, Vol. 50, pp. 225-282.
Hamilton, A. (2006), The Copts and the West 1439-1822, Oxford: Oxford University
Press.
Harby, M. K. and E. M. El-Azzawi (1960), Education in Egypt (U.A.R) in the 20
th

Century, Ministry of Education: Education Documentation Centre of U.A.R., Cairo.
Heyworth-Dunne, J. (1938), An Introduction to the History of Education in Modern
Egypt, UK: Luzac & Co.
Hütteroth, W. and K. Abdulfattah (1977), Historical Geography of Palestine,
Transjordan and Southern Syria in the Late 16th Century, Erlangen: Selbstverlag der
Fränkischen Geographischen Gesellschaft.
Ibn-Abdel-Hakam (1974), Futuhat misr wa akhbaruha (Conquest of Egypt and its
History), M. Sabeeh (Ed.), Cairo: Mu’assasat Dar Al-Ta’awun.
Ibn-Al-Gy’aan (1898), Al-tuhfa al-sanneya bi asmaa’ al-bilad al-misriya (The Treasure
Book of Names of Egyptian Localities), Cairo: Al-Matba’a Al-Ahliya.
Ismail, A. (1998), Al-nuzum al-maliya fi misr wal sham zaman salateen al-mamaleek
(Fiscal Systems in Egypt and the Levant under the Mamluk Sultans), Cairo: Egyptian
General Book Authority.
Issawi, C. (1981), “The Transformation of the Position of the Millets in the Nineteenth
Century,” In The Arab World’s Legacy by Charles Issawi, Princeton: The Darwin Press.
John of Nikiu (1916), The Chronicle of John Bishop of Nikiu, R. H. Charles (Tr.),
London: Williams & Norgate.
Karpat, K. H. (1978), “Ottoman Population Records and the Census of 1881/82- 1893,”
International Journal of Middle East Studies, Vol. 9, pp. 237-274.
Krämer, G. (1989), The Jews in Modern Egypt: 1914-1952, I.B. Tauris & Co. Ltd,
London.
Kuran, T. (2004), “The Economic Ascent of the Middle East’s Religious Minorities: The
Role of Islamic Legal Pluralism,” Journal of Legal Studies, Vol. 33, pp. 475-515.
Kuznets, S. (1960), “Economic Structure and Life of the Jews,” In The Jews: Their
History, Culture, and Religion, L. Finkelstein (Ed.), pp. 1597-1666. Philadelphia: Jewish
Publication Society of America.
186

La Porta, R., F. Lopez-de-Silanes, A. Shleifer and R. W. Vishny (1997), “Trust in Large
Organizations,” American Economic Review, Vol. 87, No.2, pp. 333-8.
Lane-Poole, S. (1969), A History of Egypt in the Middle Ages, New York: Haskell House
Publishers.
Little, D. (1976), “Coptic Conversion to Islam under the Bahri Mamluks 692-755/1293-
1354,” Bulletin of the School of Oriental and African Studies, Vol. 39(3), pp. 552-569.
Maghawri, S. (2000), Al-alqab wa asma’ al-hiraf wal-waza’if fi daw’ al-bardiyat al-
arabiya (Titles and Names of Crafts and Jobs in the Light of Arabic Papyri), Three
Volumes, Cairo: National Library Press.
Mahmoud, A. (2009a), Al-jizya fi misr 1713-1856 (The Poll Tax in Egypt 1713-1856),
Cairo: Supreme Council of Culture.
Mahmoud, A. (2009b), “Dafatir tarabee’ al-misaha (Cadastral Registers),” in Abhath
nadwat khamsun ‘aman ‘ala insha’ dar al-wathai’q al-qawmiya (Proceedings of the
Seminar on Fifty Years since the Establishment of the National Archives of Egypt), R.
Hilal et al. (Eds.), Cairo: National Archives and Library of Egypt.
Malik (1985), Al-Muwatta’, M. Abdel-Baqi (Ed.), Cairo: Al-Halabi.
Markiewicz, T. (2009), “The Church, Clerics, Monks and Credit in the Papyri,” in
Monastic Estates in Late Antiquity and Early Islamic Egypt: Ostraca, Papyri, and Essays
in Memory of Sarah Clackson, A. Boud’hors et al. (Eds.), Cincinnati: The American
Society of Papyrologists, pp. 178-204.
Marsot, A. L. A. (1984), Egypt in the Reign of Muhammad Ali, Cambridge University
Press, UK.
Mikhail, M. (2004), Egypt from Late Antiquity to Early Islam: Copts, Melkites, and
Muslims Shaping a New Society, PhD Dissertation in History, UCLA.
Mokyr, J. (2002), The Gifts of Athena: Historical Origins of the Knowledge Economy,
Princeton University Press, Princeton, NJ.
_______ (2005), “The Intellectual Origins of Modern Economic Growth,” Journal of
Economic History, Vol. 65(2), pp. 285-351.
Morimoto, K. (1981), The Fiscal Administration of Egypt in the Early Islamic Period,
Dohosha.
Newman, S. (1991), “Does Modernization Breed Ethnic Political Conflict?” World
Politics, Vol. 43(3), pp. 451-78.
187

Owen, R. (2002), The Middle East in the World Economy 1800-1914, I. B. Tauris & Co
Ltd, UK.
Pamuk, S. (2006), “Estimating Economic Growth in the Middle East since 1820,”
Journal of Economic History, Vol. 66(3), pp. 809-28.
Papaconstantinou, A. (2007), “They Shall Speak the Arabic Language and Take Pride in
it: Reconsidering the Fate of Coptic after the Arab Conquest,” Le Muséon, Vol. 120 (3-4),
pp. 273-99.
Ramzi, M. (1994), Al-qamus al-gughrafy lil bilad al-misriya min ‘ahd qudama’ al-
misriyeen ila sanat 1945 (Geographic Dictionary of Egyptian Localities from the Time of
Ancient Egyptians Until 1945), Five Volumes, Cairo: Egyptian General Book Authority.
Raymond, A. (1973), Artisans et Commerçants au Caire au XVIIIe Siècle, Two volumes,
Institut Français de Damas.
Richter, T. S. (2009), “The Cultivation of Monastic Estates in Late Antique and Early
Islamic Egypt: Some Evidence from Coptic Land Leases and Related Documents,” in
Monastic Estates in Late Antiquity and Early Islamic Egypt: Ostraca, Papyri, and Essays
in Memory of Sarah Clackson, A. Boud’hors et al. (Eds.), Cincinnati: The American
Society of Papyrologists, pp. 205-12.
Roberts, C. (1979), Manuscript, Society, and Belief in Early Christian Egypt, Oxford:
Oxford University Press.
Rogowski, R. (1985), “Causes and Varieties of Nationalism: A Rationalist Account,” In
New Nationalisms of the Developed West, Edited by: E. A. Tiryakian and R. Rogowski,
Allen & Unwi, Boston, pp. 87-108.
Ruggles, S. and R. R. Menard (1995), Public use microdata sample of the 1850 United
States census of population: User's guide and technical documentation, 7-10,
Minneapolis: Social History Research Laboratory.
Russell, J. C. (1958), Late Ancient and Medieval Population, Philadelphia: The
American Philosophical Society.
Salama, G. (1963), Tarikh al-ta’lim al-agnabi fi misr fil qarnayn al-tasi’ ‘ashar wal
‘ishrin (History of Foreign Education in Egypt in the 19th and 20th Centuries), Supreme
Council for Art, Literature, and Social Sciences, Cairo, Egypt (Arabic).
Sami, A. (1928), Taqwim Al-Neel (Chronicles of the Nile), Three volumes, Dar Al-
Kutub, Cairo.
188

Samir, S. K. (1996), “The Role of Christians in the Fatimid Government Services of
Egypt to the Reign of Al-Hafiz,” Medieval Encounters, Vol. 2(3), pp. 177- 99.
Severus (1907), History of the Patriarchs of the Coptic Church of Alexandria, Four
Volumes, B. Evetts (Tr.), In Patrologia Orientalis (Volumes I, V, and X).
Sheikho, L. (1987), Wuzara’ al-nusraneya wa kuttabuha fil islam 622-1517 (Christian
Ministers and Scribes in Islamic History 622-1517), K. Al-Yasu’yi (Ed.), Beirut: Al-
Maktaba Al-Paulisiyca.
Sijpesteijn, P. (2009), “Landholding Patterns in Early Islamic Egypt,” Journal of
Agrarian Change, Vol. 9(1), pp. 120–133.
Tagher, J. (1998), Christians in Muslim Egypt: A Historical Study of the Relations
between Copts and Muslims from 640 to 1922, Oros Verlage, Altenberge.
Toledano, E. (1998), “Social and Economic Change in the Long Nineteenth Century,” In
Cambridge History of Egypt, Vol. 2, Edited by: M. W. Daly, Cambridge University
Press.
Tousson, O. (1926), La Géographie de l'Egypte à l'Epoque Arabe, Volumes 1 and 2,
Cairo: L’Institut Français de l’Archéologie Orientale.
van Leeuwen, M. H. D. and I. Maas (2005), “A Short Note on HISCLASS,” In History of
Work Information System, http://historyofwork.iisg.nl/
Wellhausen, J. (1927), The Arab Kingdom and its Fall, M. G. Weir (Tr.), Calcutta:
University of Calcutta.
Werthmuller, K. (2010), Coptic Identity and Ayyubid Politics in Egypt 1218-1250, Cairo:
American University in Cairo Press.
Wiet, G. (1927), “Kibt,” in The Encyclopædia of Islam: A Dictionary of the Geography,
Ethnography and Biography of the Muhammadan Peoples, Vol. 2, M. Houtsma, A.
Wensinck, T. Arnold (Eds.), Leiden: Late E.J. Brill and London: Luzac.
Wilfong, T. (1998), “The Non- Muslim Communities: Christian Communities”, in The
Cambridge History of Egypt, Vol. 1, C. F. Petry (Ed.), Cambridge: Cambridge University
Press.
Yuchtman, N. (2010), “Teaching to the Tests: An Economic Analysis of Traditional and
Modern Education in Late Imperial and Republican China,” Unpublished manuscript.
Ziadeh (1915-17), “L’Apocalypse de Samuel, Supérieur de Deir-el-Qalamoun,” in Revue
de L’Orient Chrétien, Vol. 20, pp. 374-407.
189

Appendix A

Scanned Pages of the Census Registers

A.1. A Page from an Urban Province (Cairo)

Page 896 of Register of Abdin District, Cairo Province, 1848, Preserved at the National Archives of Egypt
190

A.2. A Page from a Rural Province (Al-Gharbiya)

First page of register of the Village of “Bigirim wa Kafr al-Sheikh Mansour,” Al-Gharbiya Province, 1847,
Preserved at the National Archives of Egypt

191

Appendix B

Actual Examples of the Digitized Census Records

1. (Al-Mu’allem) Antonios Luqa: Male, Free, Able-bodied, HH Head, 40 years,
Inside the government’s control, Coptic Christian, From Abu-Tig (in Abu-Tig, Asyut),
Scribe at the Customs Department in Bulaq.
Address: 15 Harat Al-Izba (From Darb Al-Geneina), Shyakhat (Quarter of) Youssef
Alam, (District of) Al-Azbakiya, (Province of) Cairo.
Type of Dwelling: House Waqf (Religious Endowment) of Copts.
2. Salman Abdel-Rehim: Male, Free, Able-Bodied, HH Head, 45 years, Inside the
government’s control, Muslim, From Tukh (in Qena, Qena), Farmer.
Address: House of Salman Abdel-Rehim, Hissat Naser Eleiwa, (Village of) Tukh,
(District of) Qena, (Province of) Qena.
3. Mardukh Youssef: Male, Free, Able-Bodied, HH Head, 25 years, Inside the
government’s control, Jew, From Cairo, Goldsmith.
Address: 35 Harat Al-Yahud Al-Qarra’een (Karaite Jews), Shyakhat (Quarter of)
Khidr Ibrahim, (District of) Bab El-Shiriyya, (Province of) Cairo.
Type of Dwelling: House owned by the heirs of Ibrahim Khidr.

192

Appendix C
Data Appendix to Chapter 2
C.1. Coding the Occupational Titles in the Egyptian 1848 and 1868
Census Samples and Constructing the Social Status Index (SSI)

The 1848 and 1868 Egyptian individual- level census records provide the earliest
comprehensive lists of occupational titles in Egypt and perhaps in the Middle East at
large. An important phase of the digitization project of the Egyptian census records (See
Chapter 1) consists of the coding of the occupational titles, to make it usable in
quantitative analysis. To this end, in the data entry phase of the digitization project,
occupational titles were first entered in full text in Arabic exactly as they appeared in the
census manuscripts. In the post- data entry phase, I coded the occupational titles in the
digitized samples manually according to Historical International Standard Classification
of Occupations (HISCO). The HISCO coding of the Egyptian occupational titles will be
made available on the web with free access to all, through the website of History of Work
Information System (http://historyofwork.iisg.nl/). In this section, I document the process
of the manual coding of the occupational titles and the construction of the Social Status
Index (SSI). In particular, I point out the challenges and difficulties of coding the
Egyptian historical occupational titles, and the decisions that I had to make in the manual
coding process. It has to be emphasized, however, that any occupational coding is by its
very nature arbitrary, and thus that this manual coding is merely one possible coding
scheme out of many possible schemes. Yet, an important advantage of the current coding
scheme is that it follows the HISCO system, and will hence make the future
193

harmonization of the Egyptian census records with the North Atlantic Population Project
(NAPP) historical census records from North Atlantic populations relatively
straightforward.
In the absence of an Egyptian dictionary of occupational titles in the nineteenth
century, I had to rely on the available historical sources on guilds and occupations in
nineteenth century Egypt (Baer 1964; Raymond 1973; Ghazaleh 1999), besides common
sense, and historians’ advice
1
in understanding the Egyptian occupational titles. This was
followed by finding the closest English (or French) translation of the Egyptian
occupational title, and then assigning to it the HISCO code of the corresponding English-
language title. HISCO scheme provides a brief description of the tasks of each
occupational code. In this regard, the HISCO coding phase can be considered as a first
step towards building a dictionary of occupational titles for nineteenth century Egypt.
Finally, I created new codes for the occupational titles for which I did not find a close
English or French counterpart (Table C.2), based on my understanding of the HISCO
major (and minor) group the title falls under. For the few titles that I could not
understand, I coded them using the five-digit scheme with the first two digits being “xx”,
followed by three-digit serial number. Fortunately, the individuals with unknown
occupations represented less than 1 percent of the population with recorded occupational
titles in 1848 and 1868. These observations are not included in the empirical analysis and
they are treated as having “missing” occupational titles.

1
I am particularly indebted to Emad Hilal at the National Archives of Egypt (NAE) for his help in this
respect.
194

Although the coding process is in principle a straightforward one, several challenges
and difficulties arose, which I summarize as follows:
1. Difficulty of distinguishing production from commercial activity in some
occupational titles: The problem mainly arises because of the peculiarity of the Arabic
language, where the occupational title is merely relating the individual to the product he
is producing/selling. For example, strictly speaking, the title “hariry” merely relates the
individual to the “silk” product and could possibly mean that the individual is
manufacturing silk or is rather merely selling silk. This causes confusion on the proper
classification of the occupational title in the HISCO scheme. Nevertheless, according to
Raymond (1973, 1:213), many of these occupational titles in fact involved both
manufacturing and trade activities, where the artisan was actually selling his products.
Hence, I chose to classify the occupational title as a “production” title if it was possible to
produce the product in a workshop in the geographical location of the individual. For
example, the title “tabban” which relates the individual to the “hay” product does not
involve a production activity since the title existed only in the cities where making hay
was not feasible, and hence I coded it as a sales occupation.
2. Omitted occupational titles where the establishment of work or employer is
mentioned: In a few cases, the census scribe does not mention the occupational title of the
individual and merely mentioned the work establishment or the employer of the
individual. In most of these cases, however, it is possible to infer the occupational title
from the information mentioned on the establishment of work. For example, an individual
may be recorded as (working) in a specific military battalion, which implies that he is a
195

“soldier”. Also, an individual might be mentioned as (working) in a government
manufactory or a workshop, where I infer that the individual is a “factory worker.” This
also applies to most of the students in the modern public schools, the religious elementary
schools kuttabs, and the higher religious institutes such as Al-Azhar, where the word
“student” is not mentioned explicitly, and only the educational establishment is
mentioned.
3. Primary and secondary jobs: In a few cases, there is more than one occupational
title mentioned for the individual because the individual has two jobs. In these cases, I
always coded the first mentioned occupational title unless one occupational title is a
specialization (or a further explanation) within the other “general” occupation. Hence, for
example an individual who is recorded as a “farmer and guard” is coded as farmer. But,
specialized military personnel, such as a lieutenant engineer, were classified according to
their specialization (engineer), regardless of the order of the occupational titles.
4. Change in occupation or in labor force participation status: In a few cases, the
individual has recently changed jobs (or has recently become unemployed) and both the
past and present jobs are recorded (with an indication of the timing). In these cases, I
always take the present occupation unless the current status is “unemployed,” where I
take the past occupational title instead. This is motivated by my interest in the occupation
the individual is generally working in rather than in his exact status at the time of the
census.
5. Vague occupational titles: For some occupational titles, I used other information
to determine the exact meaning of the title. For example, the title tabe’ (follower) is
196

coded based on the individual’s legal status. Free followers are coded as domestic
servants, but slave followers are coded as slaves. Another vague title is khaddam
(servant), which could possibly mean a free domestic servant, a slave servant, or a free
employee employed by the “master/employer.” In all these cases, I have to combine other
information on the household relationships, and legal status in order to classify the title
into one of these categories. For all the vague occupational titles, I created occupational
codes to explicitly code them such as: khaddam (employee), mustakhdim (employee),
shaghaal (manual worker), mo’awen (assistant- nonmanual), mosa’ed (assistant-
manual).
Constructing the Social Status Index (SSI)
Based on the HISCO coding of occupational titles, I assigned values for social status
standing of each occupational code following the HISCLASS measure described in van
Leeuwen and Maas (2005). HISCLASS is based on classifying occupational titles
according to four dimensions: manual versus non-manual, skill level, supervisory versus
non-supervisory role, and primary versus non-primary sector. The authors use the 1965
Dictionary of Occupational Titles (DOT) of the USA in order to define and to classify the
HISCO occupational titles/codes along the aforementioned dimensions. Table (C.1)
shows the dimensions of HISCLASS with the modification that I made by extending the
HISCLASS value of (6) to include unskilled non-manual workers. Three notes are in
order: First, the information used in the HISCO coding or in HISCLASS is the
information on occupational title only. The information on establishment of work is not
used in the coding except in the very few cases when the occupational title is itself
197

missing. Second, for the new occupational codes that I created and thus do not exist in the
HISCO scheme, I had to assign values for HISCLASS. Table (C.2) shows the decisions
that I made in this regard, which are based on my understanding of the nature and skill
content of these occupational titles (either from direct knowledge and/or from the
aforementioned historical sources). Third, for the occupational titles that exist in the
HISCO scheme, I used the original HISCLASS values that are available through the
website of History of Work Information System (http://historyofwork.iisg.nl/). In order to
construct the Social Status Index (SSI), and to account for the fact that occupations in
nineteenth century Egypt may have had quite different social class standing from the
occupations in 1965 USA, as reflected in the DOT, I revised the original HISCLASS
values, that are based on van Leeuwen and Maas (2005). This revised measure is the one
that I am using in the empirical analysis throughout chapter 2. As a robustness check,
however, I used the original HISCLASS measure that is purely based on the U.S.
classification, to replicate all the results. The results are not altered. Moreover, the
correlation coefficient between the original HISCLASS and the constructed Social Status
Index is 0.87.
C.2. Constructing the Modernization Index
I constructed the (industrial) modernization index (Modern) that takes the value of
one if the individual is employed in a state modern production project. Fortunately, the
census takers are usually keen on distinguishing between the individuals who are
employed by the state (miri) and those who are working in the private or “outside” sector
(barrani). The reason for this interest on part of the census takers lies, perhaps, in the
198

growing central power of the state over the nineteenth century. For most of the
individuals employed by the government, the name of the establishment of work was
recorded in the occupation field. I constructed a dataset of state “modern” projects, by
project’s name, industry, and location, among other information on dates of construction
and closure, number of machines, and number of workers, in both 1848 and 1868, based
on information from both secondary historical sources (M. Fahmy 1954; Al-Gritli 1952;
Sami 1928) and the digitized census samples. I then combined the list with the individual
census records in order to construct the individual-level modernization measure.
2

Perhaps, an example is helpful in illustrating the procedure of creating both the
Social Status Index and the Modern indices. Suppose an individual is recorded in Cairo’s
census manuscripts as a “carpenter in Bulaq’s arsenal.” I first enter this information in
full in Arabic, and I then code the occupational title independently from the establishment
of work. On the one hand, I manually code the occupational title “carpenter” according to
the HISCO scheme with the code “95410” which corresponds to “Carpenter, general.”
Following HISCLASS, I assign the value “6” to this title on the ladder of the SSI which
corresponds to “medium-skilled non-supervisory manual workers in the non-primary
sector.” On the other hand, based on the secondary dataset of state modern production
projects that I constructed, I know that Bulaq’s arsenal is one of the state production
projects in the military industries. Hence, this individual takes one in the modernization

2
Modern is independent from the Social Status Index (SSI). The criteria for the SSI, including primary
sector vs. non-primary, are inferred from the occupational title only. Information on work establishment
that is used to construct the modernization index is not used in constructing the SSI. Thus, a livestock
worker in a textiles manufactory is given the same SSI value as a livestock worker on a farm, although the
former is working in the modern sector while the latter is in the traditional sector.
199

index. Importantly, if the individual is recorded as a carpenter only; i.e. employed in the
traditional sector, then he would have been assigned the same occupational code and the
same SSI as the carpenter in the modern sector.
C.3. Constructing the Industrial Affiliation Variable
I constructed a categorical variable named Industry that classifies the individual into
one of five industries: Textiles, transportation, military, other modernized industries, and
non-modernized industries. The first four categories exhaust all the industries in the
modern sector, while the last category includes all the industries for which there is no
state modern project operating based on the dataset of state modern projects that I
constructed. The individuals for which the establishment of work is known (including all
the individuals in the modern sector) are classified based on the industrial affiliation of
the establishment they are working at regardless of their occupational title. The industry
of the individuals for which only the occupational title is known is inferred from their
title. For example, a weaver and a merchant of cloth in the traditional sector with no
further information mentioned are classified into the “textiles” industry. A scribe in a
textiles manufactory is classified in the “textiles” industry too. An animal-driver in the
traditional sector and a scribe in the railways in the modern sector both belong to the
“transportation” industry.
C.4. Other Variables in the Empirical Analysis
1. Religion: This is recorded directly in the 1848 census manuscripts either in front
of the individual record itself or in the tabulation that follows each street or village
section. In 1868, it is mostly unrecorded, but it is still straightforward to infer it from the
200

names. Religion in Egypt is an inherited characteristic that is decided by birth and not by
choice. It is the same as race in the U.S.
2. Age: This is recorded in years as of the next birthday. Individuals who are not
enumerated often have missing age (See Chapter 1).
3. Slave: This is a dummy variable that takes the value of one if the individual is a
slave or an emancipated slave.
4. Black: This is a dummy variable that takes the value of one if the individual is
Sudanese, Nubian, or Abyssinian.
5. Foreigner: This is a dummy variable that takes the value of one if the individual is
recorded as “Outside government’s control” or is of “Protégé” status.
6. Migrant: This is a dummy variable that takes the value of one if the individual is
recorded as born outside the province of residence.

201

Table C.1: Dimensions of HISCLASS
Manual
/ Non-
manual
Skill Supe-
rvision
Sector Class labels HISCLASS
Non-
manual
Higher skilled Yes Other Higher managers 12
Primary
No Other Higher professionals 11
Primary
Medium skilled Yes Other Lower managers 10
Primary
No Other Lower professionals, clerical and
sales personnel
9
Primary
Lower skilled Yes Other
Primary
No Other Lower clerical and sales personnel 8
Primary
Unskilled Yes Other
Primary
No Other Unskilled non-manual workers 7
Primary
Manual Higher skilled Yes Other
Primary
No Other
Primary
Medium skilled Yes Other Foremen 7
Primary
No Other Medium-skilled workers 6
Primary Farmers and fishermen 5
Lower skilled Yes Other
Primary
No Other Lower-skilled workers 4
Primary Lower-skilled farm workers 3
Unskilled Yes Other
Primary
No Other Unskilled workers 2
Primary Unskilled farm workers 1
Source: van Leeuwen and Maas (2005). I made two modifications on their scheme. First, I added the
category “unskilled non-manual workers” with the corresponding value “7” which is the same as that of
“foremen.” Second, I reversed the values of the HISCLASS index, when constructing the SSI, so that
higher values indicate higher social status.
202

Table C.2: Occupational Titles in the Egyptian Census Records that Do
Not Exist in HISCO Scheme and Their Assigned HISCO Codes and
HISCLASS Values

Code Occupational title and description Definitional notes HISCLASS
-11 Student or apprentice Out of labor force
-10 Student in a public or military school Out of labor force
-8 Child Out of labor force
-6 Student in Al-Azhar or other higher
education religious institute (mugawir)
Out of labor force
-5 Unemployed; without occupation Out of labor force
-2 Student in a religious elementary
school (kuttab)
Out of labor force
-1 Retired Out of labor force
12510 Legal representative (Wakeel) 4
12610 Petition Writer ('Ardh'algi) 6
13600 Fiqi (Kuttab teacher and/or general
religious worker)
6
13800 Higher Religious Institutes Teachers 2
17155 Singers of poems praising the Prophet 11
21250 Waqf manager 3
31010 Governmental scribe 4
33115 Banker 4
36050 Ticket conductor on trains and trams
(kumsari)
5
39910 Private scribe 4
49040 Traders in slaves (galslab and yasirji) 5
49050 Broker (simsar) 5
53260 Coffee and tea waiter (qahwaji) Serves tea and coffee
in an oriental
coffeeshop
9
54070 Slave 11
54075 Eunuch; Head of female slaves 11
54080 Emancipated slave 11
55250 Water tanks filler in public buildings
(malla’)
9
58410 Headmen (sheikhs) of villages and
urban quarters
3
59300 General-purpose assistant or servant
(khaddam)
May serve in
domestic households
or in businesses or in
public
establishments.
Chosen if product or
industry cannot be
inferred.
5

203

Table C.2. (Continued)
Code Occupational title and description Definitional notes HISCLASS
59400 Shader makers and organizers
(shawadri)
9
59500 Shoe Polishers (bouyaji) 9
59910 Water porters (saqqa) 9
62130 Farm Slaves 12
65000 Nomads, Bedouins, dwellers (‘orbaan) 12
94400 Weighters and measurers,
specialization unknown
7
95980 Railways construction workers 11
99940 Beggar 11
99950 Assistant, unspecified manual work
(Mosa'ed)
Chosen if product or
industry cannot be
inferred.
11
99960 Assistant, unspecified non-manual
work (Mo'awen)
Choose if product or
industry cannot be
inferred
5
99970 Employee, unspecified (Mostakhdim) 5
99980 Laborer or manual worker (Shaghaal) 11

Note: I reversed the HISCLASS values when constructing the SSI.

204

Appendix D

Data Appendix to Chapter 4

1. Arab Immigration Waves to Egypt in the Seventh to Ninth Centuries: The
information is based on Al-Barry’s (1992) Arab Tribes in Egypt in the First Three
Islamic Centuries, which, in turn, draws on Arabic medieval narratives. The author
traces, for almost all the immigrating tribes, their destination in Egypt mostly at the kura-
level, the administrative division that existed in 640-1036. In a few cases, the destination
of the Arab tribe is recorded at the village- or medieval province-levels. In the first case, I
matched the destination to the district that the village belongs to in the 1897 census. In
the second case, I matched the destination to the 1897-districts that comprise the
medieval province. Drawing on Ramzi’s (1994) Geographic Dictionary of Egypt and
Tousson’s (1926) La Géographie de l'Egypte à l'Epoque Arabe, I matched the kuras to
the 1897-districts, based on the location of the chef-lieu of each kura. No information on
the borders of the kuras survived. Fortunately, the average surface area of the kura was
comparable to that of the 1897 district (markaz). The number of kuras in the Nile Delta
and Valley in the seventh to ninth centuries (76) (Ramzi 1994, p. 31) was almost the
same as the number of rural districts in 1897 (81). Egypt’s inhabited area, the Nile Delta
and Valley, has hardly changed since ancient times.
2. Coptic Churches and Monasteries in the Twelfth and Fifteenth Centuries: I draw
on two medieval sources: (1) Abul-Makarim’s (1999) Churches and Monasteries of
205

Egypt provides a comprehensive list of the Christian religious establishments, Coptic,
Melkite, and Armenian, that existed in Egypt at the end of the twelfth century. There are
two versions of this book. The first is The Churches and Monasteries of Egypt and Some
Neighboring Countries that was edited by Evetts and was first published in an English
translation from the original Arabic manuscript in 1895 where it was attributed wrongly
to Abu-Saleh the Armenian. This early version included only the institutions in Nile
Valley. Fortunately, in the 1990s, Anba-Samuel published a two-volumes version of the
book in both Arabic and English. The first volume included the missing part about Nile
Delta, while the second volume was a re-publication of Evetts’ edition on Nile Valley.
The book is now believed to belong to the twelfth century Coptic chronicler Abul-
Makarim.
(2) Al-Maqrizi’s (2002) Lessons in Examining Plans and Monuments includes a less
comprehensive list of churches and monasteries (both Coptic and Melkite) in the fifteenth
century. Both lists are organized on a geographic basis at the urban street- and rural
village-levels. I matched the locations in both sources to the 1897-urban districts and
rural villages based on the streets’ and villages’ names that hardly changed over time, and
the information provided by the books’ editors in identifying the streets and/or villages.
In this matching procedure, I excluded the churches and monasteries in “extinct”
medieval locations that were deserted by 1897 based on Ramzi (1994). I then used Ramzi
(1994) and the 1477 medieval source, Ibn-Al-Gay’an’s (1898) Names of Egyptian
Villages, to exclude, from the 1897 rural villages list, the “post-1477” villages that
emerged after the Ottoman Conquest (1517) (43 percent), and thus to construct a
206

modified list of the 1897 villages that reflects the map of medieval rural Egypt.
3
In
addition, I excluded the villages that were named after Coptic monasteries in 1897 since
the vast majority of the inhabitants of these villages in 1897 are the monastic population.

3
Ibn-Al-Gay’an’s 1477-list is based on Al-Rok Al-Nassery, the computation of Egypt’s area that was
carried out under the Mamluk sultan Al-Nasser Muhammad Ibn-Qalawun in 1315. However, Ibn-al-Gay’an
updated the original list to 1477 and thus his list includes the villages that emerged between 1315 and 1477.
An improvement over this 1477-village list is possible through using the first comprehensive village list in
Egypt that was produced in Al-Rok Al-Hussamy, under Hussam-Al-Din Lajeen in 1298, and that was,
preserved without updates in a contemporary manuscript called Tuhfat Al-Irshad that presumably exists in
some library in Egypt (Ramzi 1994, pp. 18-21). I was not able, however, to find this 1298 manuscript.

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Three essays on cooperation, social interactions, and religion

PDF

Explaining well-being: essays on the socio-economic factors accompanying life satisfaction

PDF

Essays on the politics of the American welfare state

PDF

Three essays on the microeconometric analysis of the labor market

PDF

Essays on health economics

PDF

Essays in international economics

PDF

Two essays on financial crisis and banking

PDF

What leads to a happy life? Subjective well-being in Alaska, China, and Australia

PDF

Three essays on social policy: institutional development, and subjective well-being as a cause and consequence of labor market outcomes

PDF

Essays on health insurance programs and policies

PDF

Adaptation, assets, and aspiration. Three essays on the economics of subjective well-being

PDF

Essays in political economy and mechanism design

PDF

The impact of trade liberlaization on firm performance in developing countries -- new evidence from Pakistani manufacturing sector

PDF

Three essays on health & aging

PDF

Essays on development and health economics

PDF

Essays on the estimation and inference of heterogeneous treatment effects

PDF

Essays on the empirics of risk and time preferences in Indonesia

PDF

Essays on family planning policies

PDF

Essays on economics of education and private tutoring

PDF

Three essays in health economics

Asset Metadata

Creator Saleh, Mohamed (author)

Core Title Essays on the new economic history of the Middle East

School College of Letters, Arts and Sciences

Degree Doctor of Philosophy

Degree Program Economics

Publication Date 04/11/2012

Defense Date 03/20/2012

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag Conversion,Copts,Middle East,Modernization,OAI-PMH Harvest,occupational segregation,poll-tax,religious minorities

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Costa, Dora (committee chair), Nugent, Jeffrey B. (committee chair), Boustan, Leah (committee member), Easterlin, Richard A. (committee member), Miller, Donald Earl (committee member)

Creator Email msaleh@usc.edu,msaleh1982@yahoo.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-26623

Unique identifier UC11288161

Identifier usctheses-c3-26623 (legacy record id)

Legacy Identifier etd-SalehMoham-590.pdf

Dmrecord 26623

Document Type Dissertation

Rights Saleh, Mohamed

Type texts

Source 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 a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA