Human Capital Development in the Middle East: Is ...ehes.org/ghanem.pdfRima Ghanem University of Tübingen ABSTRACT Human capital and its relation with religion is an important topic

1

Human Capital Development in the Middle East: Is Secularism a Solution?

Evidence from Turkey in the 19th and 20th Century

Rima Ghanem

University of Tübingen

ABSTRACT

Human capital and its relation with religion is an important topic especially in the Middle East

region. This paper tests the influence of separating religion and government on human capital

in Turkey. The development of numeracy is calculated using evidence based on two Turkish

censuses which were carried out in the twentieth century. Was secularism the solution to

develop human capital in Turkey? Did the Turkish regions equally benefit from secularism

and the other reforms that were carried out by the Ataturk government? And was there a

significant difference between human capital development in the Kurdish and the Turkish

regions?

1. INTRODUCTION

The relationship between religion and human capital has been discussed in earlier studies.

Some researchers showed a positive relation between religion and education. For example,

Botticini and Eckstein (2007) provided evidence for a positive influence of being Jewish on

people’s education. Similarly, Becker and Woessmann (2009) showed that Protestantism has

led to better education levels in Prussia already before the industrialization. In Turkey, during

the time of Mustafa Kemal Ataturk in the first half of the 20th century, many reforms have

2

been initiated in different fields such as politics, economics and culture. Ataturk’s reforms can

be summarized mainly in abolishing the sultanate and afterwards the caliphate system in the

country and converting the republic of Turkey in a secular state. Turkey with its Muslim

majority changed from being an Islamic state to a laic country. The main hypothesis of this

research is that, separating the religion from the government had a positive influence on the

human capital in Turkey. Introducing the secular state of Turkey might have led to a clear

jump in numeracy levels in the different Turkish regions. Apparently Mustafa Kemal paid

attention to the education in the republic of Turkey. He built new schools and made primary

schooling mandatory and free. In addition to that, he replaced religious education by a

national education system. Turkey did not change to an atheist state; the freedom to worship

and follow religions still existed. But the idea was concentrating on Islam in the mosques and

religious places; on the other hand, what mattered at school was science and education.

The methodology applied in this paper is the “age heaping methodology”. It relies on

the ability of people to declare their exact ages without rounding them to numbers ending with

zero or five. A measure that represents the share of people that are able to state their age

correctly is the ABCC index. This index will be calculated for 57 Turkish provinces according

to gender and different birth decades. The data used in this research stem from censuses that

were carried out in different years in the twentieth century in Turkey and cover the birth

decades of the 1860s-1950s. These censuses contain the declared ages of people in the

Turkish provinces according to their gender. Some available resources are used to determine

the shares of religions and ethnic groups in Turkey during the same period. By using these

data sources and applying the age heaping methodology, this research will assess whether

converting to a secular state in the first half of the 20th century had a significant effect on the

human capital development in Turkey. The Turkish growth rate during and after the reforms’

3

period will be compared to other countries’ growth rates in the region. Other factors that will

be taken into account are gender inequality and ethnic influence on human capital.

2. HISTORICAL BACKGROUND

Before the establishment of the modern Turkish Republic in 1923, the Ottoman Empire was

ruled by the Islamic law. This law controlled all of the important fields in people’s lives such

as education which was in case it was available mainly religious. (Çağırıcı, 1992 cited in

Rankin and Aytaç, 2006).

On October 29th

, 1923, Turkey was announced as a republic state and Mustafa Kemal

was elected to be its first president. A few months later, the Grand National Assembly voted

to abolish the Islamic law on March 3rd

, 1924 (Davison, 1981). The caliphate system was

abolished on November 2nd

, 1924 (Cagapaty, 2006) and Turkey converted into a secular state.

The main definition of secularism is the separation of church and state (Yildirim, 2004).

Ataturk wanted not only to separate the government from religion, but also to limit the control

of religion and put everything under the government’s control (Geibel, 2012). His main goal

was to change Turkey into a modern secular state which would be independent and able to

compete with other developed nations. Therefore, he introduced reforms into different fields.

He was successful in strengthening the relations with other countries trying to avoid wars in

order to give Turkey a chance to rise up. During his fifteen years of government, Ataturk

introduced a westernized institutions system, reforms in the language, a new capital in Ankara

and many changes in the cultural field. He also initiated laws that strengthened the role of

women and encouraged them to improve their position, in addition to many other reforms.

However, these fifteen years were not enough to achieve better results in the economic field.

The Second World War afterwards affected the economy negatively. There are other negative

4

aspects to consider. For example, the majority of the rural population remained unaffected by

the modernization. Some scholars, such as Landau (1984), have criticized the reforms in that

the changes did not impact on the core of the social and economic structure; in other words,

they were superficial.

Because of its importance, the Ataturk government applied changes to the education

system. The Arabic alphabet was replaced with the Latin alphabet and the Islamic education

system was replaced with a western education style (Koc et al., 2007). Primary education

became mandatory for all children, including girls. Apparently, gender equality was an aim in

the government agenda (Rankin and Aytaç, 2006). In 1924, the Ministry of National

Education regulated the whole education system in the country. It began to change the

curriculum at elementary schools according to the needs of the established republic (Gözütok,

2003). Around a decade after, in 1936, new reforms were conducted again on the curriculum.

This new curriculum remained the same until 1948 when another wave of developments was

initiated (Binbasioglu, 1995, cited in Koc et al., 2007).

The main question of this study is: Was converting from an Islamic to a secular state

beneficial of human capital development in Turkey? In order to answer this question, we will

study the development of an indicator of human capital during the secularization period and

after it. There will be a comparison between Turkey and other countries in the area for the

same period of time, to prove that secularism had a significant effect on human capital

development.

5

3. DATA DISCRIPTION

Two data sets will be used in this study. One of them is an international comparative data set

on Turkey and its neighboring countries. The second data set provides regional evidence on

Turkey. It is based on a combination of two different Turkish censuses. The first census was

carried out in 1935. The publication of this country-wide census included the aggregate

numbers of people in almost all of the Turkish provinces by single years of age. Fifty seven

provinces were included; some of which are today divided into smaller provinces.1 In this

census there is detailed information about each province: the aggregate population (according

to gender, languages, occupations and age), the area, the religion shares, and some other

details. The documents of the 1935 census are in French and in Turkish. There is a microfilm

copy kept at the Mannheim Center for European Social Research (Mannheimer Zentrum für

Europäische Sozialforschung (MZES)) and the original documents are available in Turkey.

The second census was done in 1985 and it is available online from IPUMS International2.

This census is taken from the Turkish Statistical Institute. It includes individual observations

including age statements. Evidence is available for all of the Turkish provinces. However,

only the provinces that were mentioned in the first census are included in this study. The

sample of the 1935 census reflects 7,184,645 people aggregated according to their ages and

provinces of residence. The ages taken into account of this research are between 23 and 72

years. Covering the seven geographical Turkish regions, this data comes from 57 provinces

(see Table 1). The census sample of 1985 takes into account that, a few changes happened in

1 However, five provinces were not mentioned in this census: Yalova from Marmara region in the west, Karabük

a part of the Black Sea region in the north, Kirikkale in Central Aanatolia, Tunceli in Eastern Anatolia and

Batman in Southern Aanatolia. In addition to these provinces, there was another excluded province which is

Hatay, it was a Syrian territory back then and became Turkish only in 1939 according to an agreement between

Turkey and France which controlled Syria during that time. (Pipes, 1949).

2 Minnesota Population Center. Integrated Public Use Microdata Series, International: Version 6.1 [Machine-

readable database]. Minneapolis: University of Minnesota, 2011.

6

the second census and some of the provinces were divided into smaller ones. In this paper, the

observations of these provinces are summed up according to the 1935 administrative division.

The estimated Kurdish population share in Turkey by region and by province is taken from a

study of Multu (1996).

The internationally comparative database mentioned above contains ABCC index values for

each birth decade from 1860 until 1950 for twelve countries and was provided by Crayen and

Baten 2010; and Prayon and Baten 2013. The twelve countries are all in the region around

Turkey and they are included in the research to compare the Turkish increase of numeracy.

These countries are: Russia3, Georgia, Armenia, Azerbaijan, Iran, Iraq, Syria, Cyprus, Egypt,

Greece, Bulgaria and Romania. The values of GDP per capita for Turkey and the other 12

countries in the region were taken from the Maddison Project Database by Bolt and van

Zanden (2013). The average of the GDP per capita for each decade was calculated and

included in the regression.

4. METHODOLOGY

4.1 Age Heaping

There are many indicators used nowadays to measure the human capital in a country, such as

the total number of students that are enrolled at schools, the literacy rates, the expenditure of

the government on education, the average salaries of teachers, the ratio of teachers to students

numbers, and many other classic indicators that can be calculated using the available data in

the world. But such indicators cannot be calculated for countries where the needed data is not

available, especially in earlier time periods.

3 i.e. those regions of the Russian Empire and the Soviet Union which are situated on today’s Russian territory.

7

When there are no official registers of the literacy rates in a country, the literacy level can be

estimated using signature rates as a proxy, which can be calculated using some available

documents such as marriage registers, military and prisons records and other legal documents

(A’Hearn et al., 2009). However, significant evidence is not available for some countries.

There is another indicator for the cases where there is no schooling or literacy data available

in order to determine the human capital of a country. This indicator is the numeracy level of

the people. One might wonder why not school enrolment rates or literacy are taken to

consider the effects of the Ataturk reform. But, for Turkey no evidence is available for school

enrolment before the first cohort which benefited from the Atatürk school reforms (Benavot

and Riddle, 1988) and (Lindert, 2004). For the other Middle Eastern countries taken as

comparison sample here, only some sketchy evidence on Egypt, Iraq and Syria is available.

Literacy is only available for a cross-section of regions in 1935, plus some punctual

information before.

The numeracy level can be specified as the ability of people to calculate their ages

accurately. This measure can be determined by using registers of countries where the people

mentioned their ages, such as censuses, hospitals registers, court documents and marriage

registers. Today, by using birth certificates and other documents almost everyone can

determine his or her exact age, but this was not the case in earlier times. It has been noticed

that people tended to round their ages. They mostly reported them as numbers ending with

zero or with five. Some of the younger population reported their ages ending with multiples of

two. This tendency to report the ages ending with a particular digit is called age heaping.

Earlier literature shows that there is a positive correlation between the age heaping and

illiteracy among people. Compared to educated people, illiterate ones seem to be less able to

determine their exact ages. More age heaping was noticed for women and older people as

well.

8

4.2 ABCC Index

There are a number of methods to calculate the ratio between the heaped ages and the other

ones. According to A’Hearn, Baten and Crayen (2009), the best way is by using the Whipple

Index.

The Whipple index determines the ratio between the number of people in a census who

declared rounded ages (ending with zero or five) and the total number of people in this

census.

[∑

⁄ ∑

]

In our study, the ages used in the analysis are between 23 and 72, ages younger than 23 or

older than 72 are excluded from the study.

The result of Wh = 500 means that all of the people reported rounded ages. The other way

around, when everyone reports his or her exact age, the Whipple Index should be equal to

100. That means, twenty percent of the people reported ages ending with a multiple of five,

which is supposed to be the ideal case.

A’Hearn, Baten and Crayen (2009) suggested a new index to simplify the interpretation of age

heaping. We call it the ABCC Index4, and it is a linear transformation of the Whipple Index.

The ABCC index estimates the ratio of the people who stated their exact ages to the whole

population.

4 The name of this index comes from the initials of the authors’ names in addition to Greg Clack’s, who gave this

suggestion in a comment on the paper.

9

[

]

The ideal case of Wh =100 leads to the result of ABCC = 100, which means that all of the

people reported their ages correctly. The worst case will be when everyone reports rounded

ages and the resulted ABCC will be in this case equal to zero.

A’Hearn, Baten and Crayen (2009), also discuss a number of caveats such as potential

cultural bureaucratic and sample selection biases which we do not discuss here. Sample

selection bias is not a problem here, as the sample is based on census data.

According to A’Hearn et al. (2009), in order to calculate the ABCC index, the ages are

divided into groups according to their birth decades. The range of the ages is, as mentioned

above, between 23 and 72 and it is divided into four age groups each includes ten years, “and

two ages ending with a multiple of five”. The resulting five age groups 23-32, 33-42, 43-52,

53-62 and 63-72 are classified into five birth decades and then the ABCC index can be

estimated for each birth decade.

5. ANALYSIS

5.1 Age Heaping and ABCC Index for Turkey

The available data about Turkey indicates clear age heaping in 1935. However, this age

heaping gets less with time and shows a clear improvement in 1985. It differs between

provinces. Some provinces show better numeracy and less age heaping than others, as in the

case of Istanbul for example. A marked difference is noticed between men and women, the

10

females tend to heap their ages more than men in all of the provinces. For example, where the

women show higher age heaping than men in the Black Sea region (see Figure 1).

The ABCC Index for Turkey is calculated for the 57 provinces as well as the seven

geographical regions by gender and by birth decades from 1860 till 1950. Figure 2 shows the

development of ABCC index by birth decade for each region. There is a huge difference

between the ABCC values for the regions in the birth decade 1860. While the Marmara region

in the northwest shows the best value of more than 55%, on the other side, Eastern and

Southern Anatolia are characterized by less than half this value by less than 25%. This

difference becomes less for the younger generations and declines till around 15%. But the

ranking of the regions stays almost the same with Marmara region and Aegean region on the

top and Eastern and Western Anatolia at the lowest level.

The analysis indicates a clear gender inequality, where the female ABCCs are lower

than the male ones in all of the provinces. Figure 3 shows the values resulting from

subtracting the females ABCCs from the males ABCCs for the seven regions by birth

decades. All of the regions improved with the time and the inequality was getting less for the

youngest generations. This is not only caused by the bounded variable problem, because even

in the later periods, values of 100 were not reached. But still there is inequality by around

15% in the east and south, while it vanishes in the west for the people who were born in 1950.

The difference was not only among regions, it was also in between each region where

some provinces (the smaller regional unit) had different levels of ABCC than others in the

same region. However, these differences are still not that large in each region. In the 1935

census, one can notice that most of the provinces in the west have a good ABCC level in

comparison with the eastern provinces. The provinces in the central part have mostly middle

values (see Figure 4). The provinces’ ABCC values improve in the 1985 census. Almost all of

the provinces double their ABCCs. Figure 5 shows this improvement and it shows that the

11

eastern part of Turkey is still less developed than the western part. What could be the reason

of this difference?

5.2 Effect of Kurdish ethnicity and other factors

In the following section, we deviate from the analysis of the main question about the effect of

the Ataturk reforms in order to study the differences within Turkey.

In order to understand the differences in numeracy between the Turkish provinces and

regions, one should study the geographic and demographic differences between them. The

eastern part of Turkey is a mountain area (Oxford Business Group, 2009). This part is

inhabited mainly by the Kurds who are an ethnic group, have a common language and history

and mostly are Muslims (Multu, 1996). It is not easy to estimate the exact number of Kurds in

Turkey, even nowadays. Koc et al. (2008) estimate the percentage of Kurds to be 14% of the

whole population depending on the Turkish Demographic and Health Survey of 2003. In 1965

82% of them lived in the eastern part of Turkey. There is some other estimation according to

the old statistics of 1927 and 1935 which included the distribution of people according to their

languages. Figure 6 shows the distribution of Kurds in Turkey according to the 1927 census.

In this map the high density of Kurds is obvious in the eastern part of Turkey compared to the

other regions. Multu (1996) estimated the number of Kurds in the different provinces in 1990

relying on data from 1965 and 1990 censuses. These estimates will be used in regressions to

estimate the effect of being a province with Kurdish majority on the numeracy of the people.

The economic mechanism is partly traditional discrimination of the Kurdish ethnicity within

the Turkish dominated state; partly the more traditional economic and social system which

developed, following the language boundaries.

12

Table 2 shows the results of the OLS regressions. The dependent variable is the ABCC

value. The reference category is a female from Aegean region who was born in birth decade

1860. In the first regression the Kurds variable is included as a dummy equals one if the

majority of the province are Kurds. The result indicates a significant negative effect of being a

province of Kurdish majority. Being a male has a significant positive effect on ABCC which

supports the hypothesis of gender inequality in the different regions. The geographical regions

show significant differences, while only being a province in Marmara results in a better

ABCC level compared to the Aegean. Other regions followed in the order: Mediterranean,

Black Sea and Central Anatolia. Eastern and Southern Anatolia indicate no significant

difference in the first regression. But after excluding the Kurds dummy in the second

regression, they show significant negative effect by around 20% lower ABCC level. In the

regressions 5 and 7, the Kurds percentage share in the whole population in 1990 in each

province according to Multu’s estimates is included resulting a significant negative effect on

numeracy.

The numeracy disadvantage of the Kurdish provinces did not change during the

Ataturk reforms. During the years of War of Independence (1919-1922), Mustafa Kemal

Ataturk used to recognize the mixture of the ethnic groups in Turkey. He declared many times

that “Kurds and Turks are true brothers”. He also used to mention “the people of Turkey”

rather than “the Turkish people” while talking about his people. But on the other side, his

critics mention that after 1923 all his government’s concentration was on the cultural

revolution, and it invested little to develop the Kurdish provinces (Mango, 1999). In the same

paper, Mango states that there was resistance against the land reforms that Ataturk conducted

in the Kurdish areas. The opposition was mainly from the Kurdish landowners who refused

these reforms because they were afraid to lose their noble titles as a result. Other scholars

criticize Ataturk’s policy of ignoring the non-Turkish ethnic groups and discouraging them

13

using their languages in public. Aytürk (2011) mentions that it was not allowed to speak

Kurdish in public and the names of the places should be changed into Turkish. This might was

a reason of having less benefit from the educational reforms. Encouraging learning Turkish at

schools in addition to Kurdish might not be a bad idea. But it might have been a reason of

having less benefit from schools when only Turkish was allowed, not the mother language of

the Kurdish majorities in the east.

In addition to the effect of ethnic group, we expected a significant effect of time.

Therefore the birth decades were included in the regression, and they had significant positive

effect over time. The later the person is born, the better ABCC value he has. However, while

the effect of the birth decades grows clearly over time, between the birth decades 1930 and

1940 there is only a slight growth. It is still not yet clear why the people who were born in the

decade of World War II and after Ataturk’s death would show only a modest growth in their

ABCC level.

Another factor which might play a role in the numeracy level is the religion. The

majority of the Turkish population is Muslim (approximately 99%). Around 85% of the

Muslims are Sunni and the rest are Alevi. The Non-Muslims are divided into Jews, Catholics,

Greek Orthodox and Protestants (Oxford Business Report, 2009) and most of them live in

Istanbul. The Non-Muslim share per thousand (1935 census) is included in the regression and

it shows a small but a significant positive effect on numeracy in all of the regressions even if

the population density is controlled for. Being from a province with high density has no

significant effect on numeracy.

Other dummies are included in the regression to test whether the neighboring countries

affect the ABBC levels. Being a neighbor of Georgia or Armenia has no significant effect.

However, being a neighbor for the Muslim countries in Asia has a significant negative effect.

14

Surprisingly, being from the provinces on the European borders has an unexpected negative

effect on the numeracy.

In Regression 6, the regions are not included, but a province from each region is

included. The result is a significant large positive effect for Istanbul from Marmara region as

expected. A significant positive effect appears of Izmir from the Aegean region, Antalya from

the Mediterranean and Ankara the capital in Central Anatolia. While being from Van from

Eastern Anatolia, Trabzon from the Black Sea or Mardin from the south result a significant

negative effect on the numeracy.

5.3 Ataturk Reforms and Numeracy

In order to test the effects of the Ataturk reforms, we adopted a very straightforward approach

as a first step. We compiled a sample of numeracy growth rates for the Ataturk period, for

Turkey before the reforms, and for a set of neighboring countries. Which countries might

provide a good comparison for the territory which formed later the Turkish state? First of all,

the neighbors to the East and in the Caucasus: Armenia, Azerbaijan, Cyprus, Georgia, Iran,

Iraq, and Syria had similar numeracy levels. In contrast, Egypt was still at a lower stage of

numeracy development, but we still included this country. In order to complete the number of

neighboring countries (including those which had been ruled by the Ottoman Empire earlier),

we also included the Western and Northwestern neighbors: Bulgaria, Greece, Romania, and

Russia. We compiled evidence for these countries starting in the 1860s and running until mid-

20th

century (based on the compilations in Crayen and Baten 2010, and Prayon and Baten

2013).

As a second step, we assigned a dummy variable to Turkey in the decades in which the

Ataturk reforms should have exerted an influence. Of course, we need to control for initial

15

levels of numeracy: a country with a high level of numeracy would have more “room” for

substantial increase. In contrast, a country which has already reached a high level of basic

numeracy development can achieve only small growth rates. One additional problem is that

some of the countries, especially in the Northwest, had already reached values of 100 percent

in the last few decades. Hence we tested a separate specification to cope with this potential

problem: We included only those countries which had lower numeracy values than 95

percent. 95 percent is a value beyond which there is little space for development left.

The results indicate that the Kemalist reforms had in fact a significant and positive impact, if

the initial level is controlled for (Table 3). We use clustered standard errors to avoid potential

consequences of serial correlation.5

In addition, we controlled for other potential variables. Given the fact that some of

these variables have missing values for some of the countries, we entered these controls

separately, and not in one comprehensive model (including all at one time, N would be

reduced to 13).

(1) We include GDP per capita – although we would not state a certain direction of causality

here – because one could imagine that the growth rate depends on available financial

resources.

(2) Juif and Baten (2013) have recently assessed whether early land inequality has an

influence on educational formation. They shared their estimates with us, hence we could also

control for this variable in our small sample of Turkish neighbor countries.

(3) As a third variable, we controlled for democracy. If the participation rate is higher, we

would expect voters to demand a higher growth rate of educational investment.

5 The estimates are in fact heteroskedasticity-robust and cluster-robust, see Cameron/ Trivedi (2009), p. 85. We

thank Christina Mumme for this hint.

16

(4) Finally, a high level of fertility might reduce the propensity to increase, as the quality-

quantity trade-off might lead to lower investments if children numbers are high.

These control variables never reach statistical significance, which might be caused by the

small number of cases included in our sample. More importantly, however, those control

variables do not render the Ataturk reform variable insignificant. The Kemalist reforms

remain significant in all models test.

As we have seen above, there was a substantial variation between the various

provinces of Turkey. Hence, we were curious whether the Ataturk reform dummy would still

be significant and positive if applied to (a) predominantly Kurdish and (b) non-Kurdish

provinces. In fact, in this regression with provincial observation units, the effect of the

Kemalist reforms remained positive and significant in the non-Kurdish provinces. But the

significance disappeared in the Kurdish ones (Table 4).

6. CONCLUSION

The empirical analysis of the available data about Turkey in the twentieth century showed that

secularism helped improving the human capital. The reforms of Ataturk on the different fields

of life led to a significant improvement in the human capital of Turkey. However, the effect

was not the same for all the regions. In our descriptive analysis, we showed that the eastern

part of Turkey, where the Kurds are mostly located, never caught up fully to the west. Gender

inequality was shrinking gradually with the time after the reforms with less improvement for

the Kurdish regions.

This research contributed to understanding the relation between secularism and human

capital especially for the Middle East, where in most of the countries, the religion still

controls the governments and the human capital needs to be improved. As an outlook more

17

detailed studied on gender and ethnic inequality before and after educational reforms are

planned, as well as, extending the regional database to additional Middle Eastern countries.

REFERENCES

1. A’Hearn, B., & Crayen, D. J. Baten (2009). Quantifying Quantitative Literacy: Age

Heaping and the History of Human Capital. Journal of Economic History, 68(3), 783-

808.

2. Aytürk, İ. (2011). The Racist Critics of Atatürk and Kemalism, from the 1930s to the

1960s. Journal of Contemporary History, 46(2), 308-335.

3. Benavot, A., Riddle, P., (1988). The expansion of primary education, 1870–1940:

trends and issues. Sociology of Education 66 (3), 191–210.

4. Becker, S. and Woessmann, L. (2009). Was Weber wrong? A human capital theory of

Protestant economic history. The Quarterly Journal of Economics, 124(2), 531-596.

5. Binbasioglu, C. (1995). Türkiye’de Eğitim Bilimleri Tarihi (History of Turkish

Educational Sciences). The Series of Research and Examination. Ankara, Turkey:

MEB Publication (cited in Koc et al. 2008).

6. Botticini, M., and Eckstein, Z. (2007). From farmers to merchants, conversions and

Diaspora: Human capital and Jewish history. Journal of the European Economic

Association, 5(5), 885-926.

7. Cagapati, S. (2006). Islam, Secularism, and Nationalism in Modern Turkkey, Who is a

Turk? Routledge, Taylor and Francis Group, 13-20.

8. Çağırıcı, M. (1992). “İslam Düşüncesinde Aile Ahlakı” [Family Ethics in Islamic

Thought]. Pp. 341–52 in Sosyo-Kültürel Değişme Sürecinde Türk Ailesi, cilt I.

Ankara, Turkey: T. C. Başbakanlık Aile Araştırma Kurumu. Cited in Rankin and

Aytac (2006).

18

9. Cameron, A. C., & Trivedi, P. K. (2009). Microeconometrics using stata (Vol. 5).

College Station, TX: Stata Press. P.85

10. Crayen, D., and Baten, J. (2010). Global Trends in Numeracy 1820-1949 and its

Implicationsfor Long-Run Growth. Explorations in Economic History 47 (1): 82–99.

11. Dumont, P. “The Origins of Kemalist Ideology,” in Ataturk and the Modernization of

Turkey, ed. Jacob M. Landau (Boulder, CO: Westview, 1984): 29.

12. Davison, R. (1981). Turkey a Short History. 3rd edition 1998. The Eothen Press

England.

13. Geibel, B. (2012). Turkish Secularism: A Model or an Anomaly? Working paper.

14. Gözütök, F.D. (2003). Curriculum development in Turkey. International Handbook of

Curriculum Research, (pp. 607-622).

15. Juif, D. T., and Baten, J. (2013). On the human capital of Inca Indios before and after

the Spanish Conquest. Was there a “Pre-Colonial Legacy”? Explorations in Economic

History, 50(2), 227-241.

16. Koc, I., Hancioglu, A., and Cavlin, A. (2008). Demographic differentials and

demographic integration of Turkish and Kurdish populations in Turkey. Population

Research and Policy Review, 27(4), 447-457.

17. Koc, Y., Isiksal, M. & Bulut, S. (2007). Elementary school curriculum reform in

Turkey, International Education Journal, 8(1), 30-39.

18. Landau, J. (1984). Atatürk and the modernization of Turkey, Westview Press ; Leiden,

the Netherlands : E.J. Brill, Boulder, Colo.

19. Lindert, P. H. (2004). Growing Public: Volume 1, The Story. Cambridge University

Press.

20. Mango, A. (1999). Atatürk and the Kurds. Middle Eastern Studies, Vol. 35, No. 4, 1-

25.

19

21. Multu, S. (1996). Ethnic Kurds in Turkey: a Demofraphic Study, International

Journal of Middle East Studies, 28, 517-541.

22. Pipes, D. (1949). Syria beyond the peace process. The Washington Institute for Near

East Policy no. 40, p. 54.

23. Prayon, V. and Baten, J. (2013), Human Capital, Institutions, Settler Mortality, and

Economic Growth in Africa, Asia and the Americas,Working paper Tübingen

University.

24. Rankin, B. and Aytaç, I (2006). Gender Inequality in Schooling: The case of Turkey.

Sociology of Education, Vol.79, pp.25-43.

25. Tapper, N. and Tapper, R. (1987b). Thank God We’re Secular! Aspects of

Fundamentalism in a Turkish Town, L. Caplan (ed.) Aspects of Religious

Fundamentalism, London, pp.51-78.

26. Yildirim, S. (2004). Expanding secularism's scope: an Indian case study. Am. J.

Comp. L., 52, 901-903.

27. The Report: Turkey 2009, Oxford Business Group (2009)

Data Sources:

1. IPUMS: Minnesota Population Center. Integrated Public Use Microdata Series,

International: Version 6.1 [Machine-readable database]. Minneapolis: University of

Minnesota, 2011.

2. Maddison Project Database by Bolt and van Zanden (2013).

3. Mannheim Center for European Social Research (Mannheimer Zentrum für

Europäische Sozialforschung (MZES))

20

APPENDIX

Table 1: The 1935 census’ sample: Number of observations of people between 23 and 72

years old according to 57 provinces and seven geographical regions

Black Sea Region 1491952 Central Anatolia Region 1225813

Amasya 56659 Ankara 236456

Bolu 110460 Cankiri 76055

Coruh 110586 Eskisehir 83183

Corum 122008 Kayseri 132490

Giresun 107953 Kirsehir 59344

Gümüshane 63050 Konya 247827

Kastamonu 169128 Nigde 102595

Ordu 110309 Sivas 179842

Samsun 142725 Yozgat 108021

Sinop 82329 Tokat 128274 Trabzon 143442 Zonguldak 145029 Marmara Region 1518279 Eastern Anatolia Region 776192

Balikesir 223005 Agri 42635

Bilecik 59928 Elazig 99946

Bursa 211701 Erzincan 64747

Canakkale 106022 Erzurum 159573

Edirne 84572 Kars 118559

Istanbul 504368 Malatya 176196

Kirklareli 86237 Mus 57240

Kocaeli 152646 Van 57296

Tekirdag 89800 Aegean Region 1145810 Mediterranean Region 562802

Afyon 133816 Antalya 107065

Aydin 124750 Burdur 44706

Denizli 130633 Icel 106343

Izmir 302309 Isparta 74856

Kütahya 160323 Maras 72214

Manisa 201431 Seyhan 157618

Mugla 92548

Southern Anatolia Region

463797

Diyarbakir 95175

Gaziantep 121916

Mardin 93810

Siirt 52058

Urfa 100838

21

Table 2: OLS regression, dependent variable: ABCC index by province and birth

decade

(1) (2) (3) (4) (5) (6) (7)

OLS OLS OLS OLS OLS OLS OLS

Variable ABCC ABCC ABCC ABCC ABCC ABCC ABCC

male 22.97*** 22.97*** 34.94*** 34.94*** 22.97*** 22.97*** 22.97***

(0.575) (0.591) (0.633) (0.672) (0.591) (0.743) (1.270)

Black Sea -11.37*** -11.30*** -15.95*** -15.93*** -12.06***

(1.038) (1.068) (1.172) (1.244) (1.047)

Central -11.41*** -11.57*** -15.73*** -15.90*** -10.43***

Anatolia (1.154) (1.187) (1.307) (1.386) (1.125)

Eastern -1.478 -21.01*** -3.113 -25.34*** -16.81***

Anatolia (2.702) (1.281) (3.027) (1.545) (1.536)

Marmara 4.015*** 4.058*** 4.267*** 4.451*** 2.395**

(1.274) (1.311) (1.474) (1.564) (1.160)

Mediterranean -3.065** -6.323*** -3.964*** -7.682*** -5.765***

(1.274) (1.244) (1.446) (1.460) (1.245)

Southern -2.195 -21.59*** -3.671 -25.67*** -18.03***

Anatolia (2.816) (1.547) (3.157) (1.860) (1.873)

Birth Decade 7.537*** 7.537*** 7.537*** 7.537*** 7.537*** 7.537***

1870 (1.285) (1.322) (1.001) (1.062) (1.321) (1.661)

Birth Decade 10.96*** 10.96*** 10.96*** 10.96*** 10.96*** 10.96***

1880 (1.285) (1.322) (1.001) (1.062) (1.321) (1.661)

Birth Decade 19.44*** 19.44*** 19.44*** 19.44*** 19.44*** 19.44***

1890 (1.285) (1.322) (1.001) (1.062) (1.321) (1.661)

Birth Decade 27.80*** 27.80*** 27.80*** 27.80*** 27.80*** 27.80***

1900 (1.285) (1.322) (1.001) (1.062) (1.321) (1.661)

Birth Decade 36.42*** 36.42*** 36.42*** 36.42***

1910 (1.285) (1.322) (1.321) (1.661)

Birth Decade 43.87*** 43.87*** 43.87*** 43.87***

1920 (1.285) (1.322) (1.321) (1.661)

Birth Decade 47.15*** 47.15*** 47.15*** 47.15***

1930 (1.285) (1.322) (1.321) (1.661)

Birth Decade 47.59*** 47.59*** 47.59*** 47.59***

1940 (1.285) (1.322) (1.321) (1.661)

Birth Decade 52.36*** 52.36*** 52.36*** 52.36***

1950 (1.285) (1.322) (1.321) (1.661)

Nonmuslims' 0.0381*** 0.0381*** 0.0705*** 0.0700*** 0.0323***

Share (0.00980) (0.0101) (0.0188) (0.0199) (0.00935)

Kurds' Dummy -19.39*** -21.98***

(2.380) (2.625)

European -3.319** -3.242** -5.125*** -5.188***

Neighbour (1.585) (1.631) (1.816) (1.927)

Christian Asian -1.303 -1.283 -2.524 -2.491

Neighbour (1.629) (1.676) (1.796) (1.905)

22

Muslim Asian -5.253*** -5.305*** -3.990*** -4.066**

Neighbour (1.338) (1.376) (1.487) (1.577)

Density -0.0264 -0.0213

(0.0338) (0.0359)

Kurds' Share -0.118*** -0.285***

(0.0248) (0.0277)

Istanbul 20.36***

(2.833)

Izmir 13.57***

(2.833)

Antalya 8.253***

(2.833)

Mardin -17.50***

(2.833)

Van -15.78***

(2.833)

Trabzon -11.69***

(2.833)

Ankara 6.240**

(2.833)

Constant 33.11*** 32.86*** 30.33*** 29.83*** 34.62*** 24.73*** 57.99***

(1.351) (1.390) (1.701) (1.803) (1.261) (1.240) (0.974)

Observations 1140 1140 570 570 1140 1140 1140

R-squared 0.854 0.846 0.906 0.894 0.845 0.755 0.276

Reference category: female, Aegean Region, birth decade 1960. *, **, *** denote significance at the 10, 5, 1 percent

level, respectively. Controls for large population and area are included for regressions (3) and (4).

23

Table 3: Regressions: determinants of numeracy growth, 1860-1950 in Turkey and nine

of its neighboring countries

(1) (2) (3) (4) (5) (6) (7) (8)

Estimation OLS OLS FE RE RE RE RE RE

Included ABCC<95 GDP/c Land ineq. Democr. Fertility

ataturkreforms 0.026*** 0.022** 0.067** 0.026*** 0.061*** 0.032*** 0.041*** 0.026**

(0.000) (0.010) (0.038) (0.000) (0.000) (0.000) (0.000) (0.020)

initial -0.002*** -0.002*** -0.004** -0.002*** -0.001*** -0.002*** -0.002*** -0.002***

(0.000) (0.001) (0.047) (0.000) (0.007) (0.004) (0.000) (0.000)

gdp_pc 0.000

(0.298)

landgini1890 0.081

(0.717)

polity2 0.002

(0.148)

Fertility 0.001

(0.755)

Constant 0.245*** 0.238*** 0.396** 0.245*** 0.036 0.156 0.218*** 0.214***

(0.000) (0.000) (0.021) (0.000) (0.769) (0.358) (0.000) (0.000)

Observations 65 40 65 65 25 32 33 37

Adj. R-squared 0.16 0.050 0.053 0.93 0.42 0.96 0.62 0.98

(In Column 4-8, the Rsq- between is reported)

Table 4: The effect of Ataturk reform in the provinces of Turkey

(1) (2)

Kurdish share >=10 <10

ataturkreforms -0.032 0.059***

(0.439) (0.000)

initial -0.005*** -0.004***

(0.000) (0.000)

Constant 0.399*** 0.323***

(0.000) (0.000)

Time fixed effects YES YES

Observations 144 369

R-squared 0.630 0.692

24

Figure 1: Age heaping in Black Sea Region by gender 1935

The values on the vertical axis are the percentages of male and female population in the region

Figure 2: ABCC index development according to birth decades on the regional level

0

2

4

6

8

10

12

14

16

25 30 35 40 45 50 55 60 65 70

Males Black Sea

Females Black Sea

0

10

20

30

40

50

60

70

80

90

100

1860 1870 1880 1890 1900 1910 1920 1930 1940 1950

aegean

black_sea

central_anatolia

eastern_anatolia

marmara

mediterranean

southern_anatolia

25

Figure 3: Gender inequality by regions (ABCC males – ABCC females)

Note: People in the age groups of 63 -72 are excluded from this chart because of the small number of males in some

provinces

Figure 4: ABCC index by province 1935

0

5

10

15

20

25

30

35

40

45

50

1860 1870 1880 1890 1900 1910 1920 1930 1940 1950

aegean black_sea central_anatolia eastern_anatolia

marmara mediterranean southern_anatolia

26

Figure 5: ABCC index by province for birth decades 1900s-1950s

Figure 6: Kurds distribution in Turkey 1927

Source: Cagapty, 2006 from Büyük Atlas (1939), Data: Statistics Yearbook Vol.2 (1929)