human capital development in the middle east: evidence
TRANSCRIPT
1
Human Capital Development in the Middle East:
Evidence from Ottoman Turkey, Syria, Palestine and Iraq the 18th and 19th century
Rima Ghanem
University of Tuebingen
INTRODUCTION
There is a wide range of studies regarding the history of the Middle East. However, there is
an evident lack of applied studies of the economic history of this area, and thus, more
studies are needed to build a better understanding of the human capital development of this
region. Nevertheless, the researchers of the Economic History of the Middle East are soon
faced with the difficulty of data scarcity. Unfortunately, it can be particularly challenging to
find the required data concerning this region for different reasons. In this paper, I will
present a large dataset which is taken from documents stored in the Ottoman Archives in
Istanbul, Turkey. These data will be analyzed for the purpose of getting a better perspective
about the economic history of the Middle East during the 18th and the 19th century.
This paper will apply the “age heaping methodology” which relies on the ability of people to
declare their exact ages without rounding them to numbers ending with zero or five. A
measurement that calculates the share of people who are able to state their age correctly is
the ABCC index which reflects the estimated share of people who have minimum numerical
skills in the sample. This index will be calculated for a sample of individuals that were living
in the Middle East during the 19th century.
2
The data used in this research stem from censuses that were carried out during different
years in the 19th century in the Ottoman Empire. The sample covers several provinces in
today’s Turkey, Iraq, Syria and Palestine. It contains age statements of males who were living
in 20 different places in the Ottoman Empire and were aged between 23 and 72 years old
during the first half of the 19th century. Around eight birth decades are covered beginning
from the mid-18th century, the individuals are divided according to their religions and place
of residence. Thus, the data allow us to calculate the different ABCC index values for the
people according to their birth decades, place of residence, province and religion.
Did the Muslims have better numeracy than the Christians and Jews? Were there a
significant difference between the today’s Arab regions and the Turkish ones relating to
human capital? How big was the effect of living in a large city on the human capital of
Ottoman regions in the 19th century? And were there a significant influence of being a city
located on a trade road on the human capital of this city?
By applying the age heaping methodology, and using the mentioned dataset as well as
including other variables containing information on the distance from a caravan route and
the city size in addition to the geographical, religion and time variables, the regressions in
this paper will try to answer the former questions in order to form a better understanding
regarding the human capital development of the Middle East in the 19th century.
THE AGE HEAPING METHODOLOGY
In order to measure the development of the human capital of a country, one can use a
variety of indicators that cover a period of time. Indicators such as, the educational level, the
investment in the country’s education system, the literacy rates and some other
measurements that indicate the human capital of the country can all be utilized. However,
3
the problem that can face the researchers in this domain, is the apparent lack of data when
it comes to a historical period of time in which needed data might not even exist in order to
calculate such indicators especially for ancient times. For this reason, the age heaping
methodology was introduced to solve this problem partially. By relying on only age records,
one can form an indicator of the human capital of a country. The age heaping is the
phenomenon of rounding the people’s ages to beloved number such as multiples of five.
When the examined individuals lack sufficient education and good level of numeracy skills,
they usually find it hard to calculate their exact ages accurately, therefore, they tend to
round their ages to numbers ending with zero or five. In some countries people even tend to
choose other numbers than zero or five, motivated by particular value within their cultures.
Since the Mid-twentieth century, researchers managed to prove the age heaping
phenomenon around the world. For example, through the results of the 1901 census of
Ireland (Budd and Guinnane, 1991), in the 1950 census of the United States (Myers, 1954),
and in different countries of Latin America during the former three centuries (Manzel et al,
2012). The researchers showed as well a correlation between the age heaping and the
literacy rates. And thus ever since, this method has been effectively used to indicate the
numeracy level of the people in a specific period as a tool of measurement of the human
capital.
THE WHIPPLE INDEX AND THE ABCC INDEX
One of the most accurate measurements of the age heaping is the Whipple index (A’Hearn,
Baten and Crayen, 2009). By calculating the ratio between the number of rounded ages in a
sample to the one fifth of the total number of stated ages, this index results in a number
between 100 and 500. The index is calculated by including the age statements of the people
4
aged 23 to 72 years. The 500 result of the Whipple index indicates the lowest numeracy level
when all the population state their ages as multiples of five. However, when the Whipple
index is equal to 100, which means that all of the individuals were able to calculate their
ages accurately and they have the basic numeracy level that enables them to calculate their
ages.
(1) 𝑊ℎ = [∑ (𝑎𝑔𝑒 25 + 𝑎𝑔𝑒 30 + ⋯ + 𝑎𝑔𝑒 70)
15⁄ ∑(𝑎𝑔𝑒 23 + 𝑎𝑔𝑒 24 + 𝑎𝑔𝑒 25 + ⋯ + 𝑎𝑔𝑒 72)
] × 100
In their 2009 research, A’Hearn, Baten and Crayen introduced a linear transformation of the
Whipple index to give a simple interpretation of the age heaping. The new index was named
after the initials of their names in addition to Greg Clack’s who suggested this name. The
ABCC index is used in this paper to measure the numeracy levels of different regions in the
Ottoman Empire. This index shows the ratio of the individuals who were able to state their
ages accurately.
(2) 𝐴𝐵𝐶𝐶 = [1 −(𝑊ℎ − 100)
400] × 100 𝑖𝑓 𝑊ℎ ≥ 100 ; 𝑒𝑙𝑠𝑒 𝐴𝐵𝐶𝐶 = 100
An ideal value of the Whipple index with 100 will lead to the ideal ABCC value of 100, all the
sample individuals could calculate their ages without rounding them and they have the best
numeracy level. The other way around, when the Whipple index value is 500, the ABCC value
will be zero showing that all of the population stated ages end with zero or five. The ABCC
index shows us not only the average numeracy level of a country, but it can also be
calculated for the different age groups in the sample enabling us to differentiate between
their numeracy levels. In this paper the sample is divided into five age groups with ten years
for each, and two rounded ages. The five age groups are: 23-32, 33-42, 43-52, 53-62 and 63-
5
72. The birth decades are also calculated using the age groups and subtracting them from
the census year.
DATA DESCRIPTION
The Ottoman Archives in Istanbul keep a treasure of documents covering not only Turkey,
but also the former Ottoman states in the Middle East and Eastern Europe. The data in this
research was collected at these archives and they stem mainly from thirty five different
records. These records were parts of around fourteen censuses carried out in the different
regions of the Ottoman Empire between 1830 and 1851. Those censuses included only the
male inhabitants of the surveyed areas. The names of the male people, their ages and some
other physical characteristics such as tall, short or facial beards were surveyed and
registered in Ottoman language, which used to be the official language of the Ottoman
Empire, using the Arabic alphabet around a century before the reforms applied on the
Turkish language. Using census data enables to get rid of the potential selectivity problem of
the dataset. Thus, the collected sample is supposed to be representative to the male
population in the region.
The area covered in this dataset is now located in today’s Tukey, Iraq, Syria and Palestine.
During the nineteenth century, these countries were states of the Ottoman Empire and their
borders were overlapped. Back then they were called “Eyalets” and they were divided into
“Sanjaks”, the sanjaks themselves were, in turn, divided into “Kazas” and the kazas into
villages. The Data were collected from nine eyalets: Aydin, Bolu, Haleb, Istanbul, Marash,
Mosul, Sham and Sayda. Around twenty kazas are included in these eyalets. In this paper the
collected data is restructured geographically according to today’s borders and countries
locations and they were included into nine regions: five in Turkey and the other four in the
6
Arabic countries. The Turkish regions are divided into: Istanbul and Kocaeli located in
northwestern Turkey in Marmara region. Aydin is in the west Turkey located in the Aegean
region. Marash which is called nowadays Kahramanmarash and is located in the
Mediterranean region in the south of Turkey. In the south there is also Diyarbekir which is
now a province in the Southeastern Anatolia region. The Arabic regions are: Aleppo, which
used to be an eyalet during the Ottoman Reign covering mainly the northwestern part of
today’s Syria Aleppo city and Edleb in addition to Iskenderun which became a part of Tukey
since 1939 and Ayintab which is called today Gaziantep in the south of Turkey. The second
Syrian region is Damascus which used to be called Sham and it was an eyalet as well and it
used to cover a large part of middle and southern region of today’s Syria, some parts of
Jordan, Lebanon and Palestine. The fourth Arabic region is North Palestine covering today’s
Acre, Algalil and Nablus. In the first half of the nineteenth century, this area was a part of
Sayda eyalet which used to be located in some parts of today’s Lebanon and Palestine. The
last region is Mosul, which was an eyalet located in northern Iraq.
DATA ANALYSIS
The censuses that include the collected sample were taken throughout thirteen different
years between 1830 and 1851. Covering these twenty one years, this dataset enables us to
calculate according to the five mentioned age groups, seven birth decades after subtracting
each age group from its census year. These birth decades are the ones between 1760 and
1820 and the people who were born during those decades were between 23 and 72 years
old between 1830 and 1851.
The census records in the Ottoman Archives in Istanbul are organized not only according to
their census year and geographical location, but also according to the religion of the
7
registered people. The Muslims were included in their own records. While other religions1
were included together in registers under the classification of “Reaya Defteri” which meant
the registers that used to refer to the non-Muslims who lived in a Muslim state “Reaya” and
they had to pay the “Cizye”, which is the head tax on the non-Muslim individuals to their
Muslim government (Shaw, 1978). This classification enables us to calculate the ABCC index
for the Muslims and the non-Muslims individually and compare the results to test whether
there was a significant difference between the numeracy levels of these different religions.
Unfortunately, the non-Muslims were not clearly divided according to their distinct religions,
and thus, it is not possible to differentiate the Christians from the Jews. It might be relatively
possible to identify some of the known names into their religion, but it is not possible to
estimate all of the names’ religions. Therefore, this research will be depending on dividing
the religions into Muslims and non-Muslims.
The nine regions mentioned above are all in the today’s Middle East region, which used to
play a very important role in the international trade. However, not all of these regions have
the same commercial importance. Some of them were located directly on important early-
nineteenth century trade routes. The ports of Iskandarun and Latakia on the Mediterranean
Sea, used to link Aleppo with Europe, while the caravan route across Iraq connected Aleppo
with India. Additionally, Istanbul, Diyarbekir and Damascus were also located on caravan
routes linking the East with the West and the North with the South (Owen, 1993). Being
located on a caravan route is included in the research as a variable expected to have a
positive influence on the numeracy level of the inhabitants of the cities on these caravan
routes. This dummy variable is explained by 1 if the city is located on the trade route and 0 if
1 The non-Muslims were from different religious communities. In 1885 the main communities were: Greek
Orthodox, Armenian Gregorian, Bulgarian, Greek Catholic, Armenian Catholic, Protestant, Latin, Jew, Maronite, Frank and some other minor communities (See: Shaw, 1978).
8
it is not. The data of the caravan routes are taken from map of the Middle Eastern caravan
routes in the Middle East economy in 1800 (Owen 1993, p.48).
Age Heaping:
Applying the age heaping method, Figure (1) shows that the inhabitants of the Ottoman
States during the first half of the nineteenth century used to heap their ages to multiple of
five. However, the younger people seemed to state rounded ages less than the older ones.
Going into more details, the states which are located now in Turkey seem to heap their ages
around multiples of five less than the ones which are located now in the Arabic countries
(See Figure 2). For the youngest and the old age groups, the Arabic and the Turkish regions
seem to have a similar behavior concerning the age heaping. For the middle age group,
however, the Arabic regions seem to heap more to numbers ending with zero than the
Turkish regions, the Turks tend to heap their ages less than the Arabs within this group.
Dividing the data into two groups according to the religions, the Muslims present more
rounded ages than the non-Muslims unless for the ages of 25 and 30 (See Figure 3).
ABCC Index:
By calculating the ABCC index, the average value for the sample is 27.73. This means,
according to the studied sample, only around one third of the Ottoman Empire population
were able to calculate their ages accurately. This problem may result from the low education
level in the Empire during the first half of the nineteenth century. Bayat (2003) states that
the education institutions in this period were mainly religious schools, and the most
important problem back then, that even if there were some schools, yet there was no
coherent curriculum to be taught to the children, each teacher used to teach whatever he
knew with no educational plan settled by a governmental committee. The Turkish regions,
9
however, were doing better with an average of ABCC value equal to 33.88 while the Arabic
regions’ average ABCC value is only 16.59. The reason behind this might be including
Istanbul in the Turkish sample, which was the most important city during that time, where a
large number of educated people might immigrated to work in the official offices of the
empire or in trade and other activities. But excluding Istanbul from the dataset does not
change the difference between the Turkish and the Arabic regions. Another possible reason
is the discrimination between the regions and preferring the Turkish regions with better
public goods and education facilities than the Arabic ones. Another difference is also
measured between Muslims and non-Muslims. With an average ABCC for the Muslims in the
whole sample of 24.97, the non-Muslims used to have a better average ABCC of 33.28. In
general, the analysis shows better results for the Turks, and better results of the non-
Muslims within the ethnic groups. With a value of average ABCC equal to 11.84, the Muslim
Arabs have the smallest numeracy level, while the non-Muslim Turks exceed the value of 38
(See Figure 4).
In order to see the development of the numeracy of this region, the ABCC index is calculated
for the different birth decades between 1760 and 1820. The values begin around 10 for the
first birth decade and reach 58 for the youngest group born during the 1820s. For the first
and last birth decade, the Turks and Arabs look to have similar ABCC values, but this is not
the situation for the other five birth decades, where the Turks exceed the Arabs by around
20 points (See Figure 5). The results of the ABCC index are different according to the religion.
The values for both Muslims and non-Muslims are similar with a difference of around ten
points for all of the birth decades unless for the 1810 birth decade where surprisingly the
difference exceeds the 30 points, the non-Muslims ABCC values increase gradually. Whereas
the Muslim ABCC values stay almost stable between 10 and 30 and then jump suddenly for
10
more than forty extra points for the last birth decade (See Figure 6). This huge difference
might be influenced by the small number of observations for the Muslims born in the 1820s
(See Figure 7). The difference between the ABCC values according to the religion in the
Arabic area during the seven birth decades stay between seven and thirty eight points in the
favor of non-Muslims. But in the Turkish region, there are fluctuations of the values between
the Muslims and the non-Muslims during this period (See Figure 8). The difference seems to
be small for the first five birth decades, but it becomes 28.5 points for the group born in
1810 in the favor of non-Muslims, and then, to be dramatically changed for the next birth
decade for around forty points, but in this decade the Turkish Muslims are on the top.
However, the number of observations for the Turkish sample in the last birth decade of
1820, is low in comparison with the other birth decades, with around 1.5% of the total
number of observations of Muslims, and 9.5% of the total number of non-Muslim
observations. Even though the average ABCC value of the non-Muslims was better than the
Muslims’ value for both the Arabic and the Turkish regions. But this was not exactly the case
for each of the studied regions respectively. For two of the regions, only data for the non-
Muslims were available. These records were for Istanbul which had the best ABCC average of
around 72 and Kocaeli in the West of Turkey with an average of 37. For the other seven
regions, the data are available for both Muslims and Non-Muslims. The non-Muslims were
doing better than the Muslims in Damascus, Mosul and Diyarbekir. The other way round was
in Aleppo, North Palestine, Aydin and Marash where the Muslims seemed to have better
ABCC average results (See Figure 9). The development of ABCC index in these nine regions
according to the birth decades of the research sample is shown in Figure (10). Istanbul is on
the front like mentioned above, the ABCC value begins with 46.6 for the 1780 birth decade
and it improves for the people born in the later birth decades to reach the value of 76.7 for
11
those who were born in 1810. Kocaeli, Marash, Aleppo and Damascus’ results develop as
well with time. Diyarbekir shows only a little development for the youngest generations.
Mosul’s results stay stable for the whole period. Aydin’s results fluctuate a lot around the
average of 32. In the small sample of North Palestine, the values are omitted for less than 30
observations, but the rest of the results fluctuate between the birth decades because of the
small number of observations, which is between 37 and 106.
REGRESSION ANALYSIS
As it is shown above in the data analysis, the numeracy level of the non-Muslims was better
than the Muslims in the Ottoman Empire during the nineteenth century. The today’s Arabic
regions have a lower ABCC average value than the Turkish ones. To test whether the
differences are significant or not, and to test the influence of living in a large city or on a
caravan route on the numeracy, an OLS regression is done (See table 1). Being Muslim or not
is depending on the data sources from the Ottoman Archives in Istanbul, were the Muslims
were classified in separate registers than the other religions. The “Arab” variable is a dummy
variable equal to one if the region is located now in an Arabic country, and zero if it is in
Turkey. The variable “large” is a dummy variable for being from a large city or not. This
dummy is estimated relying on 1885 population statistics of the Ottoman Empire cited in
Shaw (1978). Istanbul, Damascus, Aleppo, Marash and Aydin are considered to be large
cities, given the value of one and zero otherwise. The population statistics were taken
around 35 years after the last census used in this research. However, it is the best available
source for the total number of the provinces’ population in the nineteenth century. The
borders and the size of the provinces would not be the same as in the sample period, but we
can assume that they were not very different. The caravan routes used to cross through the
12
Middle East passing by the main cities, the variable “caravan” is a dummy variable taking the
value of one if the caravan route passes by the province near the city. If not, it will take the
value of zero, these values were estimated relying on Owen (1993). The time dummies are
for the birth decades which were calculated by dividing the sample to age groups and then
subtracting the center of the age group from the census year. The variable “silk, cotton and
textile production” refers whether the region raised silk worms, planted cotton or produced
textile. It takes the value of one if the region specialized in any of the mentioned activities.
The data for this variable stem from the Tuebingen Atlas of the Near East (Hartmann et al,
1979). The rest of the variables are dummies for the nine regions included in the research.
The first regression resulted in a 10% significant negative result of being Muslim on the ABCC
by 7 points lower than the average, almost the same influence of being from a later Arabic
county with 5% significance level and a lower ABCC index with 10 points. Being form a large
city would increase the ABCC value by around 13.6 points this result is significant by 1%.
Being on the caravan route does not have a significant influence on the ABCC index. The
ABCC index is supposed to increase the later the birth decade is, but this increase is only
significant for the last three birth decades which begin with the nineteenth century. In the
second model, the dummy of the economic activity is added to the regression, having no
significant influence itself, but it resulted in losing significance of the negative effect of being
from an Arabic country and reduced the significance level of being from a large city as well.
The last regression shows the average ABCC of the different regions, taking Mosul as a
reference region, with the average ABCC value of around 11.29, it seems to have the lowest
ABCC value among the nine studied regions. Diyarbekir is not significantly different from
Mosul, but all of the other regions have significantly better numeracy levels. Istanbul
leading, North Palestine comes second, then Marash, Aleppo, Damascus Kocaeli and Aydin.
13
CONCLUSION
The studied dataset with more than 41,800 age statements of people used to live in
different regions in the Ottoman Empire during the first half of the nineteenth century, in
addition to the age heaping and the ABCC index methodology enabled us to have a close
insight to the human capital situation in this region. The average numeracy level of the
sample was low. By differentiating the observations by religion, the results showed a better
numeracy level of the non-Muslims in comparison with the Muslims. In general, the current
Arabic states used to have a lower numeracy level than the Turkish regions. The size of the
city had a positive effect on the numeracy of its population. But caravan routes did not have
a significant influence on the numeracy. Generally, the numeracy level was developing with
time, but the economic activities dummy had no significant influence on the ABCC index and
the studied regions showed a variety in their numeracy levels. Thanks to the age heaping
methodology, it was possible to benefit from these available data to have a better
understanding of the human capital situation of the Ottoman Empire in the nineteenth
century.
14
INDEX
Figure 1 Age heaping in percentage, the whole sample
Figure 2 Age Heaping in percentage, ethnic groups
0
2
4
6
8
10
12
14
16
25 30 35 40 45 50 55 60 65 70
0
2
4
6
8
10
12
14
16
18
20
25 30 35 40 45 50 55 60 65 70
turk arab
15
Figure 3 Age Heaping in percentage, religion
Figure 4 Average ABCC/ ethnic group/ religion
0
2
4
6
8
10
12
14
16
18
25 30 35 40 45 50 55 60 65 70
muslims non muslims
0
5
10
15
20
25
30
35
40
45
muslim not muslim muslim not muslim
arab turk
16
Figure 5 ABCC/ birth decades/ ethnic group
Figure 6 ABCC/ birth decades/ religion
0
10
20
30
40
50
60
70
1760 1770 1780 1790 1800 1810 1820
arab
turk
total abcc
0
10
20
30
40
50
60
70
1760 1770 1780 1790 1800 1810 1820
muslim
not muslim
total abcc
17
Figure 7 Number of observations/ birth decades/ religion
Figure 8 ABCC development/ birth decades/ ethnic group/ religion
0
1000
2000
3000
4000
5000
6000
7000
8000
9000
1760 1770 1780 1790 1800 1810 1820
obs muslims
obs non-muslims
0
10
20
30
40
50
60
70
80
90
100
1760 1770 1780 1790 1800 1810 1820
arab muslim arab not muslim turk muslim turk not muslim
18
Figure 9 Average ABCC/ ethnic group/ religion/ regions
Figure 10 The ABCC development/ birth decades/ regions
0
10
20
30
40
50
60
70
80
ale
pp
o
dam
ascu
s
mo
sul
no
rth
pal
est
ine
ayd
in
diy
arb
eki
r
mar
as
ista
nb
ul
koca
eli
arab turk
muslim
not muslim
0
10
20
30
40
50
60
70
80
90
1760 1770 1780 1790 1800 1810 1820
aleppo
aydin
damascus
diyarbekir
maras
kocaeli
istanbul
mosul
north palestine
19
ABCC OLS(1) OLS(2) OLS(3)
Muslim -7.332* -7.403*
Standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1
(1)&(2)reference category: non-Muslim, Turk, born in 1760.
(3) reference category: Mosul
(4.337) (4.368)
Arab -10.24** -7.801
(4.651) (6.409)
Large 13.61*** 11.90**
(4.744) (5.677)
Caravan -0.484 3.599
(4.298) (8.518)
Bdec 1770 7.146 7.380
(10.32) (10.31)
Bdec 1780 8.901 9.024
(9.36) (9.256)
Bdec 1790 11.11 11.05
(9.195) (9.256)
Bdec 1800 19.03** 18.58**
(9.056) (9.153)
Bdec 1810 28.83*** 28.14***
(9.668) (9.809)
Bdec 1820 42.89*** 41.44***
(10.37) (11.35)
Silk, cotton, textile-
-6.032
production
(10.84)
Aleppo
32.03***
(9.407)
Damascus
22.28***
(7.542)
north Palestine
38.88***
(10.47)
Istanbul
53.05***
(9.407)
Kocaeli
18.69**
(8.709)
Aydin
18.09**
(7.306)
Marash
37.32***
(7.111)
Diyarbekir
3.173
(6.947)
Constant 14.54* 16.80* 11.29**
(8.367) (9.348) (5.028)
Observations 64 64 64
R-squared 0.501 0.504 0.538
Table 1 OLS regression, ABCC index/geographical/ time/ ethnic/ religion/ location
20
BIBLIOGRAPHY
1. A'Hearn, B., Baten, J., & Crayen, D. (2009). Quantifying quantitative literacy: Age heaping and the history of human capital. The Journal of Economic History, 69(03), 783-808.
2. Bayat, F. (2003). Dirasat fi tarih al-árab fi l’ahd al-uthmani. Dar Almada. Beirut. “Studies in the Arabic history during the Ottoman Era – in Arabic”
3. Budd, J. W., & Guinnane, T. (1991). Intentional age-misreporting, age-heaping, and the 1908 Old Age Pensions Act in Ireland. Population Studies, 45(3), 497-518.
4. Crayen, D., & Baten, J. (2010). Global trends in numeracy 1820–1949 and its implications for long-term growth. Explorations in Economic History, 47(1), 82-99
5. Hartmann, K. P., Scholz, F., & Schweizer, G. (1979). Tübinger Atlas des Vorderen Orients. Reichert.
6. Manzel, K., Baten, J., & Stolz, Y. (2012). Convergence and divergence of numeracy: the development of age heaping in Latin America from the seventeenth to the twentieth century1. The Economic History Review, 65(3), 932-960.
7. Myers, R. J. (1954). Accuracy of age reporting in the 1950 United States census. Journal of the American Statistical Association, 49(268), 826-831.
8. Owen, R. (1993). The Middle East in the world economy, 1800-1914. IB Tauris.
9. Shaw, S. J. (1978). The Ottoman census system and population, 1831–1914. International Journal of Middle East Studies, 9(03), 325-338.