Spatial Analysis of Regional Inequalities in Turkey

Abstract

In this paper, we examine regional inequalities in Turkey not only at the inter-provincial level but for three different regional definitions as well. The motivation draws on the findings of Gezici and Hewings(2001) that raised questions about inequalities not only between regions (inter-regional) but inequalities within each region. Hence, one contribution of this paper is to test the effects of aggregation and scale on the identification of regional inequalities. Spatial data analysis is employed to examine the spatial dimension of inequality in Turkey and the relationship between spatial dependence and the dynamics of growth. The results indicate that overall inequalities are decreasing, however spatial dependence is becoming more dominant. The Theil index indicates that interregional inequalities are increasing while intra-regional inequalities are declining for all spatial partitions from 1980 to 1997. Most developed provinces are enhancing overall inequalities, although there is some evidence of a spread effect on their neighbors.

Full text

Page 1

43rd Congress of European Regional Science
27-30 August, 2003
Jyvaskyla, Finland







Spatial Analysis of Regional Inequalities in
Turkey

















Ferhan GEZICI
Istanbul Technical University, Dept. of Urban and Regional Planning, Istanbul, Turkey
Fax: 090 212 251 48 95
(gezicif@itu.edu.tr)
Geoffrey J.D.HEWINGS
Regional Economics Applications Laboratory, University of Illinois, USA
(hewings@uiuc.edu)

Page 2

2
Spatial Analysis of Regional Inequalities in
Turkey1


Ferhan GEZICI
Geoffrey J.D.HEWINGS


Abstract

In this paper, we examine regional inequalities in Turkey not only at the inter-provincial level but for
three different regional definitions as well. The motivation draws on the findings of Gezici and Hewings
(2001) that raised questions about inequalities not only between regions (inter-regional) but inequalities
within each region. Hence, one contribution of this paper is to test the effects of aggregation and scale on
the identification of regional inequalities using currently accepted spatial analytic methods. The results
indicate that overall inequalities are decreasing, however spatial dependence is becoming more dominant.
The Theil index indicates that interregional inequalities are increasing while intra-regional inequalities
are declining for all spatial partitions from 1980 to 1997. Most developed provinces are enhancing overall
inequalities, although there is some evidence of a spread effect on their neighbors.
Key words: interregional inequalities, intra-regional inequalities, spatial dependence,
spatial data analysis


1. Introduction
There has been a resurgence of interest in regional disparities and inequalities as new
developments in methodology have opened the way to more creative consideration of
the problem. Since not all parts of a country have the same characteristics with respect
to resource orientation, manpower, economic, social and political history, spatial
interactions between regions and geographical location play an important role in
explaining the economic performance of regions. However, the inequality literature has
generally neglected the spatial dimension (Rey, 2001). In Turkey, the persistence of a
spatial dualism between east and west from the past until present was revealed in Gezici
and Hewings (2001), while the European Union has north and south spatial regimes (Le
Gallo and Ertur, 2001; Baumont and et al., 2001), Italy still has historical north and

1 Special thanks to Eduardo Almeida, Suahasil Nazara and Sandy Dall’erba (Regional Economics
Applications Laboratory, University of Illinois).

Page 3

3
south dualism (Mauro and Podrecca, 1994), furthermore Greece has two main regions
as Athens and non-Athens (Siriopoulos and Asteriou, 1998). This geographical
disparity within several countries is further evidence that space continues to matter,
even though the sources of disparities might be different from country to country.
Empirical studies to explore and explain these issues are needed and recent advances in
spatial data analysis not only facilitate consideration of the spatial issues of inequalities
but enhance the reliability of the empirical work as well (Goodchild, 1987).
Exploratory Spatial Data Analysis (ESDA) focuses explicitly on spatial effects and
consists of techniques to describe spatial distributions, discover patterns of spatial
association (spatial clustering), identify atypical locations (spatial outliers) (Anselin and
Bao, 1997). Recent empirical works by Rey and Montouri (1999), Rey (2001), Le
Gallo and Ertur (2001), Baumont, et al., (2001), Ying (2001) are some examples that
focus on regional inequalities and spatial dependence of growth using ESDA.
In the next section, the motivation and expectations are presented, while in the third
section the methodology and data are reviewed. In the fourth section, the analysis
focuses on the inequalities between and within region of different partitioning in Turkey
by using Theil index. In the fifth section, attention is directed to spatial dependence,
global and local clustering through Exploratory Spatial Data Analysis. Furthermore, the
spatial pattern of GDP regional growth is examined in relation to their initial level of
GDP per capita. The paper’s conclusion reviews the findings.

2. Motivation and Expectations
Regional inequality in Turkey is major issue in terms of regional policy. The analysis
of Gezici and Hewings (2001) revealed that the dispersion of GDP per capita across the
provinces and functional regions have similar trends with little evidence of
convergence. The growth rate and initial levels are essentially uncorrelated across the
provinces and functional regions resulting in the rejection of β convergence for the
1980-97 period (see Barro and Sala-i-Martin, 1995 for a review of the various types of
convergence). Further, GDP per capita is not randomly distributed, but highly clustered
and spatially dependent at the regional level (functional regions).
In the case of Turkey, one of the main goals has been maximizing national growth and
enhancing strong economic factors in order to enable the country to survive in a

Page 4

4
competitive world. The externally oriented policies and the focus on the European
Union have concentrated the privileges in the metropolitan cities, especially Istanbul,
generating significant advantages for them in the context of globalization. Policy
conflicts can be revealed between those that have, on the one hand, stimulated the
concentration of the fastest growing activities in the 1980s in large cities and a few
developed regions while others, on the other hand, have addressed development in the
poorer regions. These policy conflicts have neutralized many attempts to reduce
regional disparities and have sustained core-periphery disparities. The result of the
spatial correlation analysis provides evidence that the disparities between east and west
regions remain (see Gezici and Hewings, 2001).
In this paper, the level of the spatial analysis of regional inequalities in Turkey is
developed not only at the inter-provincial level but for three different regional
definitions as well. This need, to explore alternative geographies was generated by the
findings of Gezici and Hewings (2001), wherein there appeared to be a need to examine
not only inequalities between regions (inter-regional) but inequality within each region
(intra-regional) as well. Hence, the paper affords a limited opportunity to test the
effects of aggregation and scale on the identification of regional inequalities. Finally,
spatial data analysis offers the opportunity to include explicitly the spatial dimension in
inequality studies in Turkey and provides the option to explore the relationship between
spatial dependence and the dynamics of growth.

3. Methodology and Data
One of the main purposes of this paper is to examine the inequality not only over time,
but across regions and within regions as well. Furthermore the spatial dependence of
growth and its relationship to regional inequality in terms of GDP per capita is also
examined. Essentially, the initial question posed is how the provinces are clustering in
space in terms of growth and inequalities.
In literature, several empirical studies on regional inequalities using by the Theil Index
have focused on interregional inequalities, but in order to realize the dynamics and the
role of regions or smaller spatial units on inequalities, attention should also be directed
as well to intra-regional inequalities. The Theil index accounts both for inter-regional
and intra-regional inequalities and is presented as:

Page 5

5
() ∑
=






==
n
i
i
i
i
x
y
yxyTT
1
log:
(1)
where = population of province i relative to the national population and =GDP of
province i relative to the national GDP. By using the Theil index, interregional and
intra-regional disparities can be estimated as:
ix
iy
(
∑
=
g
∑
=
i
+






=
n
gg
n
i
i
i
xyTY
x
y
yT
11
:log
) (2)
where the left side is the Theil index measuring the disparity between regions (inter-
regional), and Y is the region g’s share of total GDP, and T
g
is the Theil index
measuring the disparities among provinces (intra-regional or within) in region g.
):(xy
g
However, there is no formal administrative regional unit in Turkey; several studies on
regional issues have used the geographical regions. Furthermore, the State Planning
Organization and State Statistics Institute have used provincial data to define 16
functional regions in 1982, but this division did not become a common aggregation for
either the empirical studies or regional policy initiatives.
In this paper, three alternative partitions are explored in order to analyze inequality from
different levels and perspectives: geographical regions, functional regions and coastal-
interior regions. Therefore, the role of spatial scale and its impact on inequality can be
examined in parallel to the way Rey (2001) adopted for his study in the US. In order to
test spatial dependence, the well-known Moran-I and Moran Scatter-plot (Anselin,
1988; 1995) were used. Moran’s I provides an indicator for spatial autocorrelation, here
interpreted to imply value similarity with locational similarity. A positive
autocorrelation occurs when similar values for the random variable are clustered
together in space and vice versa (Cliff and Ord,1981; Upton and Fingleton,1985). The
spatial dependence (global spatial autocorrelation) measure of Moran’s I is represented
by equation 3:
∑
∑∑
i
=
2
i
j
j
iij
i
z
zzw
s
n
I (3)
n is the number of regions, and are log of per capita income of each region,
are the elements of weight matrix W(n x n) and it is equal to 1 if i and j are neighbors
iz
jz
ij
w