Anda di halaman 1dari 16

RURDS Vol. 28, No. 1, March 2016 doi: 10.1111/rurd.

12043

REGIONAL HUMAN CAPITAL DISTRIBUTION AND


DISPARITIES IN TURKEY

Umut Erdem

Middle East Technical University, Department of Urban Policy Planning and Local
Governments, Ankara, Turkey

Human capital is a major driver of regional growth and provides insight into degrees of
regional disparity (Barro & Lee 1993). Addressing these disparities is crucial to economic
stability and cohesion in Turkey; therefore, we investigated the regional distribution of
human capital and its evolution between 2008 and 2012. We used district level address-
based population registration system data from 957 districts to analyze regional human
capital disparities and report significant results. Specifically, regional disparities in human
capital are decreasing in Turkey, but enormous regional district level disparities persist. A
disparity exists between the western and eastern regions in human capital at both primary
and secondary educational levels. However, this east–west dualism almost disappears with
regard to human capital at the tertiary educational level, which is heavily concentrated in
Turkey’s major cities (Istanbul, Ankara, Izmir, Kocaeli, and Bursa).

1. Introduction

As human capital is essential to economic growth, its effects are widely discussed in the
literature (Nelson & Phelps 1966; Lucas 1988; Grosman & Helpman 1991; Barro & Sala-i-
Martin, 1992; Mankiw et al. 1990; Benhabib & Spiegel 1994, Romer 1994; Aghion et al. 1998).
Romer (1994) states that long-term growth can be achieved through endogenous
improvements and innovations in technology, assuming that technological progress can be
realized in three ways. First, technological development can result unintentionally. For
instance, operators or mechanics can improve their skills through tacit knowledge and practice,
even though they may not be motivated to develop or invent new technology (Arrow 1962).
Second, developing the previous argument, technological development results from
intentionally undertaking professional research and development. Romer assumes that a
portion of the population works as researchers developing new technologies, a process that
leads to technological improvements as an output of knowledge. Third, technology spillovers
can also improve technology level, whether intentional or unintentional. In all cases, human
capital is primarily responsible for enhancing labor productivity, intensifying capital, and
achieving growth (Polasek et al. 2010).
Unlike Romer, Lucas’s (1988) model emphasizes the role of investment in education to
build human capital. He argues that self-sustaining economic growth can be achieved by

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 17

investing in human capital. His model summarizes the process as capital accumulation
stimulating increased production, which, in turn, creates new resources that can be used to
invest in human capital, such as universities or training programs. This results in increased
productivity gains and, thus, further output growth. The economy will achieve positive long-
term growth dependent on externalities in the form of investments in human capital.
However, growth is not always equal across regions within countries, and most countries
suffer unequal distribution of growth (Islam 2003; Magrini 2004; Rodrıguez-Pose 2012).
Similarly, human capital is not randomly distributed across regions within countries, but rather
is heavily concentrated in major regions and cities (Rodrıguez-Pose & Tselios 2011).
The literature contains three branches of reasoning for the spatial concentration of human
capital (Storper & Scott, 2009). The first branch is Florida’s (2002) creative class theory, which
assumes that human capital generates and forms creative environments that both stimulate
local development and growth and attract creative and talented people. This leads to a virtuous
cycle that encourages further growth. Second, Glaeser (2005) argues that climate and skills
contribute to the spatial proximity of human capital. Glaeser suggests that human capital favors
locations that offer a high quality and low cost living environment, together with a supply of
skilled workers. In the third branch, Clark et al. (2002) suggest that amenity factors, such as
museums, shopping malls, theaters, art galleries, and orchestras, are essential for the spatial
accumulation of human capital (Storper & Scott 2009).
Several amenities are clearly involved in the location choices and spatial proximity of
human capital (Storper & Scott, 2009). These amenities include low cost and high quality
living and working space, and the availability of local talent. In a virtuous cycle, cities or
regions that offer such amenities attract further human capital based on their high quality but
low cost living environment. In this regard, the geographical distribution of human capital and
the evolution of regional disparities in human capital are widely discussed in the literature.
Rodrıguez-Pose & Tselios (2011) and Caragliu et al. (2014) investigated regions in Europe. In
research with a country-level focus, McHenry (2014) studied regions within the United States,
Lopez-Rodrıguez et al. (2011) looked at Romania, Kerimoglu and Karahasan (2011) examined
Spain, Karahasan and Uyar (2009) investigated Turkey, and Heckmann (2002) focused on
China.
Given that Turkey suffers enormous regional inequalities in income and human
capital – despite development plans intended to reduce these regional differences (Gezici &
Hewings 2004) – a crucial question for Turkey is: “Do people move to jobs, or do jobs move to
people” (Storper & Scott 2009, p. 147)? Numerous researchers have examined the influences
on these regional inequalities, including research by Filiztekin (1998) and Gezici and Hewings
(2004) on income differences; Akg€ung€or (2003) on sectorial inequalities; Tansel and Gungor
S
(2000) and Celebio glu and Dall’erba (2010) on educational level; Ocal and Yıldırım (2010) on
terrorism; and Karaca (2004) on public investments. Several studies, such as that by Karahasan
and Uyar (2009), also sought to investigate the regional distribution and evolution of human
capital disparities in Turkey at the provincial level.
Human capital plays a key role in growth. Essentially, more human capital means more
growth because human capital increases labor productivity and, thus, intensifies capital (Lucas
1988; Romer 1994; Polasek et al. 2010). Human capital also affects regional technology
diffusion capacity, as regions with access to highly educated and skilled human capital can
benefit more from technology diffusion. The unequal distribution of human capital is crucial
to understanding unequal regional distribution of growth. To better understand regional

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
18 Erdem, Human capital distribution in Turkey

disparities in Turkey, this study investigates the extent of disparities in human capital among
districts of Turkey (i.e. the number of educated people relative to the total population), whether
such disparities are spatially dependent, and whether such disparities are decreasing.

2. Data and Methods

Annual address-based population registration system data (population census results)


were used to analyze regional disparities in human capital across the 957 districts of Turkey
during 2008–2012. The data set comprises 4785 (5 years  957) observations. A lack of district
level data on earlier years limited analysis to the period 2008–2012.
The disparities in human capital across the districts of Turkey are measured for different
levels of human capital, namely primary, secondary, and tertiary levels of education. Human
capital level is measured using the ratio of the educated population relative to the total
population for each district. Most empirical studies have used education to measure human
capital. While in reality, human capital clearly involves more than just education (Moretti
2003), the lack of proper data for Turkey necessitates that this study also use educational level
to measure human capital. District level education data in Turkey assigns individuals to one of
eight educational levels: illiterate, literate but did not graduate, primary school graduate, junior
high school graduate, high school or vocational school graduate, higher education graduate,
masters graduate, and doctoral graduate. From 2008, Turkstat have published annual data at
district level, including individuals older than six years of age. Turkstat education data are re-
grouped to yield a simple scheme that classifies individual educational level into primary,
secondary, and tertiary (Rodriguez-Pose & Tselios, 2011). The illiterate population is omitted
and, therefore, excluded from our analysis. The sum of primary, secondary, and tertiary level
human capital is less than the educated population of the region, because the illiterate
population is excluded. The primary educated population are assigned to the primary level,
graduates from junior high school and high school or vocational school are assigned to the
secondary level, and all those who have undertaken higher education, including masters and
doctoral graduates, are assigned to the tertiary level.
Explanatory Spatial Data Analyses (ESDA) methodology is used for detecting the spatial
relations of the distribution of human capital across the districts of Turkey. Global Moran I
analyses are applied to all human capital levels to test whether human capital is randomly
distributed across different regions or exhibits spatial correlations in particular hubs. Local
Moran I tests are also applied to all human capital levels to better understand the spatial
correlation characteristics of human capital, including whether hubs differ from their
surrounding regions. Regional inequalities in human capital are measured by the ratio of the
standard deviation of all human capital levels, namely primary, secondary, and tertiary, to their
mean. Additionally, convergence trends of human capital levels are estimated and captured by
the two forms of Spatial Durbin Model, the spatial lag and spatial error models.
Spatial weight matrices of inverse distance, 50 neighbors, and rook continuity are used for
quantifying and measuring the spatial relations and structures in the data. Spatial relations vary
according to the type of matrices, whether they are variable or binary. Variable matrices (such
as inverse distance) range from 0–1, closer neighbors are weighted higher, and the weights
decrease with distance. Binary matrices (such as 50 neighbors and rook contiguity) take 1 to
neighbors or 0 to all others, according to their target (Anselin 1988).

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 19

3. Are there Large Human Capital Disparities?

Regional disparities are higher in developing than in developed countries and have been
widely discussed in the literature (Rodrıguez-Pose 2012). Fujita et al. (2004) investigated the
dualism between coastal and interior regions of China, while Gezici and Hewings (2004)
investigated the spatiality of regional disparities and highlighted the dualism between the
eastern and western regions of Turkey. Geographic characteristics, such as coastal boundaries
and proximity to transportation nodes, and non-geographic characteristics, such as population,
financial and human capital, and innovation capacity, are essential to spatial distribution and
the evolution of regional disparities. Throughout history, differences between backward and
developed regions have been reshaped by regional conflicts, competitions, and trans-
formations, as well as by a lack of ports, geographical advantages (such as the availability of
natural recourses), sectorial characteristics, and lack of transformation capacity and dynamics.
However, developed regions have the capacity to transform their local characteristics, innovate
and progress their technology, and invest in and support their human capital.
Regional disparities have a long history in Turkey. After the establishment of the new
republic in 1923, Turkey turned to the Western world. Thus, western regions have attracted the
majority of investment and subsidies, because of their spatial proximity to the West and their
coastal location. Major cities, such as Izmir, Istanbul, and Ankara became the nation’s
industrial and financial cores. Although attempts have been made to balance this unequal
distribution of industrialization and financial development by establishing state-owned
enterprises in Eastern Turkey, these efforts have failed to achieve or sustain balanced
development between the country’s eastern and western regions (Pamuk 2014).
As a result of these factors, east–west dualism has become deeply evident in Turkey,
apparent in income, education, human capital, health, and investment sectors (SCelebioglu &
Dall’erba 2010). In 1980, Turkey initiated rapid and uncontrolled trade liberalization that
ultimately saw it enter into a customs union with Europe. This accommodation of uncontrolled
trade mostly affected domestic production, increasing regional disparities (Pamuk 2014).
Regional disparities in Turkey are widely discussed in the literature, which universally
supports east–west dualism, whereby the western regions are more developed than the eastern
in terms of income, education, health, and infrastructure. In other words, disparities in these
sectors are not randomly distributed across the different regions of Turkey, but rather are
concentrated spatially such that the east suffers relative underdevelopment. The regional
differences between Eastern and Western Turkey have recently increased and deepened,
mostly because of the greater attractiveness of the western regions to production networks. The
western regions also attract investment attention because of their spatial proximity to Europe,
in addition to other historical factors.
In line with the discussion thus far and to better understand the spatial distributions of
human capital (primary, secondary, and tertiary levels) across the regions (districts) of Turkey,
choropleth maps are used to investigate the spatial distributions of all levels of human capital
(Fig. 1). The intervals used in the maps are adjusted to obtain five equal intervals. The darker
shading denotes higher human capital. The maps show a clear east–west dualism, with higher
human capital in the western regions than the eastern for both the primary and secondary levels.
However, this dualism is not as clear for the tertiary level; tertiary level human capital was
concentrated in specific hubs in certain western regions (especially around the Marmara Sea
and in Ankara and Izmir provinces), rather than simply exhibiting an east–west dualism.

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
20 Erdem, Human capital distribution in Turkey

Figure 1. Regional distribution of human capital across the districts of Turkey

Analyzing the spatial distribution of human capital (primary, secondary, and tertiary
educational levels) from the perspective of east–west dualism can help us better understand the
degree of disparities, their geographical sense, and their interaction with space. At a cursory
glance, the spatial distribution of human capital confirms the existence of this dualism in
Turkey, but the dualism fully reflects the distribution of human capital of all levels (primary,
secondary, and tertiary). The sharpest and most obvious distinction is in the spatial distribution
of primary level human capital, which was concentrated in the western regions in both 2008
and 2012. Notably, no districts in Western Turkey have low overall human capital; in other
words, all of the districts with low overall human capital are situated in Eastern Turkey. In both
2008 and 2012, overall human capital level exhibited a spatial trend, whereby it decreased in
the east and increased in the west.
The spatial distribution pattern of secondary level human capital shares similarities with
that of the primary level. However, unlike primary level human capital, the secondary level is
decreasing from Western to Eastern Turkey while simultaneously exhibiting random spatial
proximity. In contrast to primary level human capital, for which all of the low-scoring districts
were located in Eastern Turkey, Western Turkey had districts with low secondary level human
capital in both 2008 and 2012. Moreover, the spatial distribution of districts with higher human
capital is more heterogeneous in the western than in the eastern region. The western region,
thus, also contains large differences in terms of the spatial distribution of secondary level
human capital versus that of primary level.
It is also possible to observe the east–west dualism in the spatial distribution of tertiary
level human capital, but its spatial distribution differs from that of the primary and secondary
level. Tertiary level human capital is concentrated in a few hubs and in districts of large
cities, and its spatial distribution correlates strongly with spatial proximity. It was mostly

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 21

concentrated in districts bordering the Marmara Sea, and in districts of cities like Istanbul,
Ankara, Izmir, Kocaeli, EskiSs ehir, and Antalya. Although the level of human capital decreases
from west to east, most western regions have only a low level of tertiary level human capital.
Although the trend is not as clear as for primary and secondary level human capital, tertiary
level also decreases from the west to the east of Turkey.
The ratio for primary level human capital at the district level ranges between 0.07 and 0.57
in 2008, and 0.15 and 0.61 in 2012, meaning that the district with the highest primary level
human capital had 8.15 times the amount of such capital than the district with the lowest level
in 2008, and this gap decreased to 4.2 times in 2012. The decrease in the ratio of the
discrepancy in human capital between the districts with the highest and lowest levels of
primary level human capital during 2008–2012 illuminates the convergence occurring across
Turkey in terms of primary level human capital.
The spatial distribution of secondary level human capital displays a different pattern of
east–west dualism than primary level. Specifically, while the districts with the highest
concentrations of primary and secondary level human capital are concentrated in Western
Turkey, so too are those with the lowest concentrations of secondary level human capital. The
choropleth map for 2008 shows a district level ratio of secondary level human capital ranging
from 0.02–0.35, while in 2012 it shows a range of 0.02–0.40. In 2008, the district with the
highest concentration of secondary level human capital had more than 16 times the
concentration of such capital than the district with the lowest concentration. In 2012, this gap
between the richest and poorest districts narrowed to more than 14 times.
The district level ratios of tertiary level human capital ranged between 0.003 and 0.296 in
2008, and between 0.009 and 0.342 in 2012, meaning that the district with the highest
concentration of tertiary level human capital in 2008 had 98.3 times capital than the district
with the lowest concentration. In 2012, this gap narrowed to 38 times. Thus, in 2008, Turkey
suffered enormous regional disparities in tertiary educated human capital, and despite rapidly
decreasing, these disparities remained very high in 2012.

4. Are the Disparities in Human Capital Spatially Dependent?

To test the spatial autocorrelation and dependency of the spatial distribution of human
capital, the Global Moran I test is applied individually to all categories of human capital. The
Global Moran’s I statistic is expressed as (Rey and Montouri, 1999):
Pn Pn
  i¼1 j¼1 wi;j xi;t xj;t
n
It ¼ Pn Pn ; ð1Þ
so i¼1 j¼1 xi;t xj;t

where i and j are regions; W is a raw standardized spatial weight matrix; and wij denotes 1 if
regions i and j are neighbors, and 0 otherwise. Moreover, xi,t is the log of human capital in year
t; n is the number of regions; and so is the sum of all wij (Rey and Montouri, 1999).
The Global Moran I test assesses whether the regional distribution of human capital is
random or exhibits spatial correlations and proximity patterns (Anselin 1988). The results are
presented in Table 1. The Global Moran I test yields higher percentages and significant results
at the 1% level for all human capital levels, meaning the distribution of human capital is

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
22 Erdem, Human capital distribution in Turkey

Table 1. Global Moran I test results for the regional distribution of human capital during 2008–2012
(957 districts)
Inverse distance 50 Neighbors Rook contiguity
 
P2008 0,473 0.653 0.778
S2008 0,777 0.292 0.430
T2008 0,366 0.212 0.277
P2012 0,593 0.491 0.605
S2012 0,695 0.303 0.437
T2012 0,366 0.214 0.274

significantly and highly space dependent and regionally concentrated. This also implies the
rejection of the null hypothesis of the Moran I test, namely, that human capital is randomly
distributed across the regions.
Table 1 shows the results of the Moran I test for all levels of human capital in both 2008
and 2012 for the spatial weight matrices inverse distance, 50 neighbors, and rook contiguity.
Secondary level human capital has the highest Moran I score, which means that secondary
level human capital is not randomly dispersed across the regions but rather is concentrated
around particular hubs. This characteristic distinguishes secondary level human capital from
both the tertiary and primary levels, although it differs more from the primary than from the
tertiary. However, the Moran I test also yields higher scores, which are significant at the 1%
level, indicating a strong spatial dependency in the regional distribution of different levels of
human capital.
For clarity and to better understand the local characteristics of such spatial dependencies,
the Local Moran I test is applied to all levels of human capital and yields significant results for
the characteristics of spatial dependencies. These results are persistent and complimentary to
the interpretations of Figure 1. Local Moran I statistics are expressed as:
 
xi Xn
Ii;t ¼ w x ;
j¼1 i;j i;t
ð2Þ
mo

with
Xn
mo ¼ i
x2i;t :

Besides the Global Moran I expression, mo equals the sum of the elements of wij, (Rey &
Montouri 1999). Figure 2 displays the results of the Local Moran I test for all human capital
categories in both 2008 and 2012. HH denotes high human capital districts that are themselves
surrounded by other high human capital districts; HL refers to high human capital districts
surrounded by low human capital districts; LH refers to low human capital districts surrounded
by high human capital districts; and LL indicates low human capital districts surrounded by
other low human capital districts.
For 2008, the Local Moran I test for primary level human capital reveals that LL districts
are concentrated in Eastern Turkey, whereas HH districts are concentrated in Western Turkey
but also occur in Istanbul, Izmir, and Ankara. Secondary level human capital exhibits similar

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 23

Figure 2. Local Moran I maps of regional human capital disparities across the districts of Turkey

east–west dualism to primary level, but is mostly concentrated in Turkey’s main industrial
hubs, such as the districts around the Marmara Sea, Ankara, Izmir, and EskiSs ehir, KırSs ehir,
Isparta, and Antalya. HH districts for secondary level human capital are also concentrated in
Trabzon, Artvin, and Tunceli in Eastern Turkey. These provinces are not production hubs, but
have higher than average levels of educational attainment. More specifically, HH districts for
secondary level human capital were concentrated on the shores of the Marmara Sea where they
formed a ribbon running from Tekirdag to Bursa. In terms of spatial distribution, a ribbon also
forms from Ankara to Kırıkkale for secondary level human capital.
HH, high human capital districts surrounded by other high human capital districts; HL,
high human capital districts surrounded by low human capital districts; LH, low human capital
districts surrounded by high human capital districts; LL, low human capital districts
surrounded by other low human capital districts.
The Local Moran I results for tertiary level human capital differ from those for the primary
and secondary levels, but still resemble the spatial distribution of the Local Moran I results of
secondary level human capital. Similar to the spatial distribution of secondary level human
capital, HH districts for tertiary level human capital are concentrated on the Marmara Sea, and
around Izmir, Antalya, and Ankara. Unlike secondary level human capital, HH districts for
tertiary level human capital are not concentrated in central Anatolia, with the exceptions of
Ankara and Denizli. Meanwhile, the LL districts for tertiary level human capital are located in
Eastern Turkey, particularly Diyarbakır Province.
The 2012 spatial distribution of the Local Moran I results for primary level human capital
is consistent with that for 2008. East–west dualism is also persistent. The HH districts were
concentrated in Turkey’s western region and the LL districts in its eastern region, especially in
the southeast. Similar to the spatial distribution characteristics of primary level human capital

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
24 Erdem, Human capital distribution in Turkey

in 2008, the districts of Istanbul, Izmir, and Ankara also have some LL districts of primary level
secondary level huhuman capital. The spatial distribution of secondary level human capital in
2012 differs little from that in 2008. HH man capital districts are mostly located in the western
part of the country, while the LL districts are primarily located in the eastern. The ribbons
formed around the Marmara Sea, and by Ankara and Kırıkkale, persist in 2012. Other HH hubs
for secondary level human capital are the districts of Antalya, Izmir, and EskiSs ehir in the west
and Tunceli, Trabzon, and Artvin in the east.
The Local Moran I dispersion of tertiary level human capital in 2012 resembles that in
2008. The HH districts are concentrated around the Marmara Sea, and around Izmir, Antalya,
and Ankara. However, the dispersion of the Local Moran I test for tertiary level human capital
clearly indicates the emergence of HL districts in Western Turkey. Additionally, the Local
Moran I dispersion of tertiary level human capital in 2012 clearly displays the emergence of
new LL districts in Eastern Turkey.
Table 2 displays the number of districts according to the Local Moran I test of human
capital levels in both 2008 and 2012. The table shows a transformation in the spatial
dependency patterns for all levels of human capital. The number of primary level HH districts
in Turkey decreased between 2008 and 2012, while numbers of secondary and tertiary level
HH districts increased. Additionally, increases in the numbers of secondary and tertiary level
HL and LL districts exceeded the number of HH districts. These patterns suggest that the
spatial dependency and proximity of human capital are increasing for secondary and tertiary
levels and decreasing for the primary level.
However, Table 2 also reveals that the spatial proximity and dependency of LL and LH
districts is increasing for all human capital levels, which indicates an increasingly unequal
spatial concentration of human capital at the district level in Turkey.

5. Are Human Capital Disparities Converging?

To better understand the unequal distribution of human capital of all levels across the 957
districts of Turkey, measurements were taken using the coefficient of variation (CV). The
inequality index, the CV, is expressed as:
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Xn ðyi;ty Þ2
t
i¼1 n1
CV ¼ ; ð3Þ
yt

Table 2. Local Moran I dispersion of human capital levels in 2008 and 2012
2008 2012

Primary Secondary Tertiary Primary Secondary Tertiary


High–high 147 126 69 128 128 71
High–low 2 1 1 2 2 4
Low–high 0 3 7 4 3 6
Low–low 134 100 3 142 105 10
Not significant 674 727 877 681 719 866
N 957 957 957 957 957 957

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 25

where yi,t is human capital in region i at time t; y t is its cross sectional mean; and n is the number
of regions. A higher CV indicates larger inequalities and more dispersed distribution of human
capital across regions. In this regard, Figure 3 shows the evolution of disparities in human
capital across the districts of Turkey between 2008 and 2012.
Figure 3 clearly shows that regional human capital disparities tend to decrease for all
educational levels; in other words, the disparities across regions in levels of human capital are
reducing. However, disparities across districts remain high in 2012, particularly for the tertiary
level, followed by the secondary and primary levels. The CV value for regional human capital
disparity (tertiary level) is 0.84 in 2008 and 0.65 in 2012. Moreover, the value of the regional
human capital disparity for secondary level human capital, which is lower than for tertiary level
human capital, is 0.40 in 2008 and 0.38 in 2012. Finally, the value of CV for the regional
disparity for primary level human capital is 0.24 in 2008 and 0.20 in 2012.
To be clear and to test the CV results for the evolution of human capital differences, the
Spatial Durbin Model is applied for estimating the convergence trends of human capital levels
(Fig. 3). Two forms (as in Eq. 4), with spatial dependence in the dependent variable and in the
error terms, are used. The model consists of 3828 observations (4 years  957 districts).
 
hci;t
ln ¼ g þ T̶1 ln hci;t1 þ mi;t ð4Þ
hci;t1
 
hci;t
rWln þ mi;t and mi;t ¼ lWmj;t
hci;t1

i ¼ 1; . . . ; 957 t ¼ 2009; . . . ; 2012

The dependent variable is the annual growth rate of human capital in region i at year t. The
independent variable is the initial human capital, hci;t1 :T̶1 , capturing the convergence/
divergence trend so that a negative and significant T̶1 may indicate evidence of a convergence

Figure 3. Regional human capital disparities in Turkey, 2008–2012 (957 districts)

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
26 Erdem, Human capital distribution in Turkey

pattern, which means that human capital increased in human capital poor districts at a faster
rate than in human capital rich districts (Barro & Sala-i-Martin 1992).
The model is applied to all human capital levels one by one. Table 3 represents the
estimation results of the Spatial Durbin Model. The rho captures the external growth in region i
created by its neighboring (adjacent) regions, and, the lambda captures the spatial dependence
in the error terms of i and j. Inverse distance weighing is used as the spatial weighing
methodology (Anselin 1988).
All results indicate negative coefficients significant at %1, which indicate that human
capital disparities in all levels have converging trends and the gaps in human capital are
decreasing across the districts. Tertiary level human capital has the highest coefficient, and
primary and secondary level human capital has lower and closed coefficients. The results
appear to be quite consistent with the previous findings of both CV and estimations regarding
the trends of convergence in all human capital levels.
With regard to the spatial spillovers of human capital, all human capital levels have
positive and significant spatial spillover effects, both in the spatial lag and spatial error models
(rho and lambda have positive and significant coefficients at %1 level).
Table 4 displays the dynamic distribution of human capital levels between 2008 and 2012
to show the roots, severity, and extent of human capital inequalities in Turkey and captures the
dynamic changes in all human capital levels by investigating whether inequalities in a district
remain stable, decrease, or increase between 2008 and 2012. The regional distribution of
human capital at all levels is divided into five equal intervals, ranging from I–V, as in Figure 1.
A lower interval means low human capital, and vice versa. In other words, the intervals in
Table 4 refer to the intervals determined in Figure 1.
In terms of primary level human capital, most districts are grouped in intervals III and IV.
For primary level, the number of districts rising to the higher intervals is lower than the number
falling to the lower intervals between 2008 and 2012. Similarly, for secondary level, most
districts are grouped in intervals II and III, in both 2008 and 2012. The number of secondary
level human capital districts rising to the higher intervals is lower than the number falling to the
lower intervals.
The regional disparities of human capital across Turkey are quite persistent for tertiary
level human capital; many districts (790 districts, Table 4) occupy the lowest interval for this
level of human capital in both 2008 and 2012. This shows the unbalanced regional distribution
of tertiary level human capital. However, the table shows a lack of districts falling to the lower
intervals, which means there is an upward direction of change, but such change becomes
increasingly infrequent as districts progress toward the higher intervals.
The dynamic distribution of human capital levels reveals that primary level human capital
is dispersed more widely than secondary and tertiary, being present in many districts in the

Table 3. Spatial panel estimations of regional human capital disparities in Turkey


Variables Primary Secondary Tertiary
 
a 0.0306142 0.0128083 0.0221481
ln human capital 0.0396780 0.0334722 0.0582023
rho 0.630012 0.943158 0.961443
lambda 0.353480 0.931885 0.92545
N 3828 3828 3828

denotes significance at 1 %,  at 5 %,  at %10.

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 27

Table 4. Dynamic distribution of regional human capital

Primary 2012

I II III IV V
I 20 33 1 0 0
II 9 67 9 0 0
Primary 2008 III 0 57 270 22 0
IV 0 0 116 273 1
V 0 0 0 43 36

Secondary 2012

I II III IV V

I 131 15 0 0 0
II 11 398 20 0 0
Secondary 2008 III 0 38 235 6 0
IV 0 0 37 51 0
V 0 0 0 14 1

Tertiary 2012

I II III IV V

I 790 78 0 0 0
II 0 60 12 0 0
Tertiary 2008 III 0 0 9 3 0
IV 0 0 0 3 0
V 0 0 0 0 2

highest interval. In contrast, tertiary level human capital is dispersed heavily across the districts
in the lowest interval. The number of districts with high tertiary level human capital in the
second and third intervals more than doubled between 2008 and 2012. Similar to the inequality
analyses in Figure 3, Table 4 shows that inequalities in tertiary level are decreasing as tertiary
level increases in districts with low human capital. However, the spatial distribution of tertiary
level remains strongly unbalanced and unequal across Turkey.
Figure 4 shows the Kernel Density distribution of human capital across Turkey. In
accordance with Table 4, Figure 4 reveals that primary and secondary level human capital is
distributed across Turkey more equally than tertiary level.
The unequal distribution of tertiary level human capital decreased from 2008 to 2012, but
the distribution of qualified and highly skilled human capital remained concentrated in several
hubs in Western Turkey. Figure 2 shows the tertiary level hubs in Western Turkey. The hub
cities are Aydın, Denizli, Balıkesir, and Afyon. These are known as “priority development
provinces” and are supported by the Turkish government through several economic policies
and subsidy programs. Essentially, Figure 4 displays the extent of regional human capital
disparities in Turkey.

6. Discussion

The spatial distribution of human capital provides insight into the level and distribution of
inequalities. In this regard, unequal distribution of human capital or the accumulation of human

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
28 Erdem, Human capital distribution in Turkey

Figure 4. Kernel density distribution of regional human capital in Turkey

capital in several hubs is clearly evident in Turkey (Figs 1, 2). The disparity among districts in
terms of the equal distribution of human capital is quite high, especially at the tertiary level.
The analyses reveal a heavy accumulation of all levels of human capital in Western Turkey.
Disparities in human capital are extremely high but decreasing (Fig. 3); findings consistent
with those of Karahasan and Uyar (2009).
The east–west dualism varies across the three levels of human capital and are clear for both
primary and secondary level, but less so for tertiary level. Human capital is concentrated
around Turkey’s industrial cores, which are primarily located in its western region. The level of
human capital is lower in Eastern Turkey. In general, tertiary level human capital is not
randomly distributed across all districts of Turkey, but is heavily concentrated in several major
hubs, the districts around the Marmara Sea, and the newly emerging industrial core in Central
Anatolia (Fig. 2). The majority of these are located in Western Turkey, indicating a strong
geographic disparity.
The spatial distribution patterns of human capital reveal that it is mostly concentrated in
the districts bordering the Marmara Sea, from Tekirdag to Bursa, and in the districts around the
Izmir Gulf. Human capital is also accumulated at hubs like Ankara and EskiSs ehir in central
Anatolia, and in some districts of Antalya and Mugla in Southern Turkey.
The analyses and estimations also reveal that the disparity across the districts of Turkey in
terms of human capital is decreasing, yet, disparities still existing in human capital levels. The
disparity across districts in terms of primary and secondary level human capital slightly
decreased between 2008 and 2012. Meanwhile, the disparity decreased at a higher rate for
tertiary level human capital. Although tertiary level human capital converged faster than
primary and secondary level, the disparity across districts in terms of tertiary level human
capital remains high. Human capital has a positive spillover effect, which means that an
increase in a region positively affects neighboring regions.
Empirical analyses reveal that, from the perspective of the unequal distribution of growth
across regions of Turkey, a regional convergence trend exists, which is also consistent with

studies by Filiztekin (1998), Gezici and Hewings (2004), and KılıSc arslan and Ozata gan (2009).
The unequal distribution of human capital correlates with the unequal regional distribution of

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 29

growth within Turkey in general; however, more specifically, the patterns of human capital
reveal diffusion from major cities to newly emerging industrial cores.

7. Conclusion

This study addresses the geography and evolution of disparities in the distribution of
human capital across Turkey. It reveals that while disparities among districts in terms of human
capital are decreasing, they nevertheless remain high. Primary and secondary level human
capital is slightly decreasing, while district level gaps in tertiary level human capital are rapidly
decreasing. Estimations show that human capital is positively diffusing to the neighboring
regions, which has mostly diffused from hubs and cores with a heavy accumulation of human
capital to newly emerging industrial cores in central and western Anatolia, a trend that can be
interpreted as human capital following jobs. The spatial scale is quite important in regional
studies and the east–west dualism in the spatial distribution of human capital varies according
to its level in Turkey. When we look at the spatial distribution of tertiary level human capital, it
is obvious that the east–west dualism frequently stressed in the literature is becoming blurred.
In this regard, district level analyses reveal that disparity and the uneven distribution of human
capital across the western regions are also quite high. With the exception of the Istanbul and
Ankara regions, most of the western districts share similar human capital disparities with the
eastern districts even though they are very closer to the production hubs. Thus, in the line with
the discussion so far, it is important to note that such overaccumulation of human capital in the
_
particular hubs of Turkey, such as Istanbul, _
Ankara, and Izmir, should be encouraged by
districts with low level human capital by supplying new job opportunities and new investments
in their facilities for providing and sustaining cohesion and stability.

Acknowledgment

The author thanks to Hasan Engin Duran, Hasan SarptaSs , Guldem Ozata
gan and Nuri
Yavan, and referees for their comments and contributions.

Submitted 21 August 2015.


Final version received 25 October 2015.


Send correspondence to Umut Erdem: umtherdem@gmail.com

References

Aghion, P., Howitt, P. and Garcıa-Pe~


nalosa, C. 1998. Endogenous Growth Theory. MIT press.
Akg€ung€or, S. 2003. Exploring Regional Specialization in Turkey’s Manufacturing Industry. In Regional
Studies Association International Conference. Pisa, Italy, April 12–15.
Anselin, L. 1988. Spatial Econometrics: Methods and Models. Studies in Operational Regional Science,
Volume 4, Springer, Netherlands.
Arrow, K.J. 1962. The economic implications of learning by doing. The Review of Economic Studies,
155–173.
Barro, R.J. and Lee, J.W. 1993. International Comparisons of Educational Attainment. Journal of
Monetary Economics 32(3), 363–394.

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
30 Erdem, Human capital distribution in Turkey

Barro, R.J. and Sala-i-Martin, X. 1992. Public Finance in Models of Economic Growth. The Review of
Economic Studies 59(4), 645–661.
Benhabib, J. and Spiegel, M.M. 1994. The Role of Human Capital in Economic Development Evidence
from Aggregate Cross-country Data. Journal of Monetary Economics 34(2), 143–173.
Celebioglu, F. and Dall’erba, S. 2010. Spatial Disparities across the Regions of Turkey: An Exploratory
Spatial Data Analysis. The Annals of Regional Science 45(2), 379–400.
Clark, T.N., Lloyd, R., Wong, K.K. and Jain, P. 2002. Amenities Drive Urban Growth. Journal of Urban
Affairs 24, 493–515.
Filiztekin, A. 1998. Convergence Across Industries And Provinces in Turkey. KoSc University. [Cited 08
Dec 2015]. Available from: http://myweb.sabanciuniv.edu/alpayf/files/2010/04/turkconv981.pdf.
Florida, R. (2002). The Economic Geography of Talent. Annals of the Association of American
Geographers 92(4), 743–755.
Fujita, M., Mori, T., Henderson, J.V. and Kanemoto, Y. 2004. Spatial distribution of economic activities
in Japan and China. Handbook of Regional and Urban Economics 4, 2911–2977.
Gezici, F. and Hewings, G.J. 2004. Regional Convergence and the Economic Performance of Peripheral
Areas in Turkey. Review of Urban & Regional Development Studies 16(2), 113–132.
Glaeser, E.L. 2005. Urban Colossus: Why is New York America’s Largest City? (No. w11398). National
Bureau of Economic Research, Massachusetts.
Grossman, G.M. and Helpman, E. 1991. Trade, Knowledge Spillovers, And Growth. European
Economic Review 35(2), 517–526.
Heckman, J.J. 2002. China’s Investment in Human Capital (No. w9296). National Bureau of Economic
Research, Massachusetts.
Islam, N. 2003. What Have we Learnt from the Convergence Debate? Journal of Economic Surveys
17(3), 309–362.
Karaca, O. 2004. T€ urkiye'deb€olgelerarasıgelirfarklılıkları: yakınsamavarmı? T€urkiyeEkonomi
Kurumutartısmametni 2004/7.
Karahasan, B.C. and Uyar, E. 2009. Spatial Distribution of Education and Regional Inequalities in
Turkey. [Cited 08 Dec 2015]. Available from: http://mpra.ub.uni-muenchen.de/30130/
Kerimoglu, E. and Karahasan, B.C. 2012. Location Patterns of Creative Capital and Regional Disparities
in Spain. IREA–Working Papers, Barcelona. 2012, IR12/002.
Kılıcaslan Y. and Ozatagan G. 2007. Impact of Relative Population Change on Regional Income
Convergence: Evidence from Turkey. Review of Urban and Regional Development Studies 19(3):
210–223.
Lopez-Rodrıguez, J., Fai~na, A. and Cosmin-Gabriel, B. 2011. Economic Remoteness and
Wage Disparities in Romania. Tijdschrift voor economische en sociale geografie 102(5), 594–
606.
Lucas, R.E. 1988. On the Mechanics of Economic Development. Econometric Society Monographs
29, 61–70.
Magrini, S. 2004. Regional (di) Convergence. Handbook of Regional and Urban Economics
4, 2741–2796.
Mankiw, N.G., Romer, D. and Weil, D.N. 1990. A Contribution to the Empirics of Economic Growth
(No. w3541). National Bureau of Economic Research, Massachusetts.
McHenry, P. 2014. The Geographic Distribution of Human Capital: Measurement of Contributing
Mechanisms. Journal of Regional Science 54(2), 215–248.
Moretti, M.M. 2003. The Affect Regulation Checklist. Unpublished Measure and Data. Simon Fraser
University, British Columbia, Canada.
Nelson, R.R. and Phelps, E. S. 1966. Investment in Humans, Technological Diffusion, and Economic
Growth. The American Economic Review 56(1/2), 69–75.

Ocal, N. and Yildirim, J. (2010). Regional Effects of Terrorism on Economic Growth in Turkey: A
Geographically Weighted Regression Approach. Journal of Peace Research 47(4), 1–13.
_
Pamuk, S. 2014. T€urkiye'nin 200 YıllıkIktisadiTarihi. _s BankasıK€ult€urYayınları.
T€urkiyeIS
Polasek, W., Schwarzbauer, W. and Sellner, R. 2010. Human Capital and Regional Growth in

Switzerland (No. 250). Reihe Okonomie/Economics Series, Institut f€ur H€ohere Studien (IHS),
Vienna.

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd
Erdem, Human capital distribution in Turkey 31

Rey, S.J. and Montouri, B.D. 1999. US Regional Income Convergence: A Spatial Econometric
Perspective. Regional Studies 33(2), 143–156.
Rodrıguez-Pose, A. 2012. Trade and Regional Inequality. Economic Geography 88(2), 109–136.
Rodrıguez-Pose, A. and Tselios, V. 2011. Mapping the European Regional Educational Distribution.
European Urban and Regional Studies 18(4), 358–374.
Romer, P.M. 1994. The Origins of Endogenous Growth. The Journal of Economic Perspectives 8(1),
3–22.
Storper, M. and Scott, A. J. (2009). Rethinking Human Capital, Creativity and urban Growth. Journal of
Economic Geography 9, lbn052.
Tansel, A. and Gungor, A.D. 2000. Provincial Inequalities in School Enrollments in Turkey. Economic
Research Forum Working Paper (No. 2003). Economic Research Forum, Cairo. [Cited 08 Dec 2015].
Available from: http://ssrn.com/abstract=266168 or http://dx.doi.org/10.2139/ssrn.266168

© 2016 The Applied Regional Science Conference (ARSC) and John Wiley & Sons Australia, Ltd

Anda mungkin juga menyukai