Spatial Income Inequality in Turkey and the Impact of

Spatial Income Inequality in Turkey
and the Impact of Internal Migration
by
Süleyman Özmucur*
and
Jacques Silber**
* The University of Pennsylvania.
** Department of Economics, Bar-Ilan University, Ramat-Gan 52900,Israel.
April 2002
Not to be quoted without the authors’ permission
I.
Introduction:
Almost fifty years ago, in his Presidential Address to the American Economic
Association, Kuznets (1955) suggested that income inequality was generally rising in
the early stages of economic development but declining in the latter phases of the
development process. In Kuznets’ words: “One might thus assume a long swing in the
inequality characterizing the secular income structure; widening in the early phases of
economic growth when the transition from the pre-industrial to the industrial
civilization is more rapid; becoming stabilized for a while; and then narrowing in the
latter phases.” (Kuznets, 1955, page 18). Such an inverted U relationship between
inequality and development has since been known as the Kuznets Curve and has
become one of the most famous hypotheses in economics. Kuznets (1955) had in fact
centered his argument on the impact of rural to urban migration flows on the
distribution of incomes during the development process, the idea being that “even if
within-sector inequality is constant and the ratio of mean sectoral incomes is also
constant, the shift of population between sectors at first produces a widening in
inequality and then a narrowing" (Adelman and Robinson, 1989). Such a result was
derived mathematically in subsequent work by Robinson (1976), Knight (1976),
Fields (1979) and Anand and Kanbur (1993).
Fields (1980) considerably extended this approach by making a distinction between a
sector enlargement effect, a sector enrichment effect and an interaction terms and
there have been numerous empirical investigations testing Kuznets' conjecture (see,
Fields, 2001, for a thorough survey of all these studies). Today the consensus seems to
be that “the Kuznets curve is not a necessary feature in the data, nor even the best
general description of changes over time….Other variables (than growth) also determine
inequality. They include the basic nature of the economic system itself; the structure of
output; the composition of exports; regional patterns; the structure of employment; the
distribution of land and capital; the state of development of the capital market; the level
and inequality in the distribution of human capital; and the distribution of social
income.” (Fields, 2001, pages 69-70).
The present study is not another attempt to check the validity of Kuznets' thesis,
though it emphasizes a central feature of his story: the internal migration from rural to
urban areas. The main theme of this paper is different because its focus is on the impact
of internal migration on spatial inequality. Another point to be stressed is that the
empirical illustration presented here is based not on a cross-section of countries but on
several income surveys that were conducted in Turkey, mainly on the 1987 and 1994
surveys.
Several studies have actually attempted to analyze changes in the distribution of
income in Turkey that took place in recent years. An interesting survey of individual
income distribution in Turkey is presented in Gürsel et al. (2000). Their study
determines the exact impact of various income sources on overall income inequality,
using Shorrocks (1982 and 1983) approach., examines the impact of various
household characteristics on poverty levels and finally compares the results with those
available for other EU countries. Selim and McKay (2000) propose a more detailed
analysis of the relationship between households’ heads characteristics and poverty
levels but their study is mainly descriptive. This is also generally the case of Özmucur
and Silber’s (2000) study of inequality in Turkey in 1994. In this paper they attempted
to determine the impact that various income sources (income from the primary job,
from a secondary job and from other sources) and different population categories
(Wage and Salary Earners, Daily Workers and Proprietors) in both urban and rural
areas had on the overall level of income inequality in Turkey in 1994.
To understand the changes that occurred between 1987 and 1994 in the distribution of
income in Turkey it is in fact necessary to remember what the macroeconomic
environment was during this period. The 1980’s were a period of commodity trade
liberalization where the output composition of the country was significantly modified.
This induced a reduction of the employment generation capacity of the Turkish
economy and important income redistribution effects. Moreover between 1988 and
1994 the Turkish governments implemented expansionary macroeconomic policies
financed largely through domestic borrowing at high interest rates and such policies
evidently resulted in transferring incomes between groups. In fact populist policies
and the consequent fiscal deficits of the 1990s resulted in the breakdown of the
financial system in 1994. It should be clear that such important changes must have
had also an impact on regional inequality.
As mentioned previously the main topic of this paper is to analyze the impact that
internal migration in Turkey had on spatial inequality. Such an effect may be assumed
to take place through various routes. First, because the types of employment and the
relative importance of various income sources are not the same in rural and urban areas,
internal migration affects inequality because it modifies the structure of the labor force
and the composition of income. Second because the average size of households is
smaller in urban than in rural areas, internal migration is likely to have an effect on the
inequality of per capita or of standardized (per adult equivalent) income. Third internal
migration modifies the relative importance of the various regions and this will also have
an impact on spatial inequality.
The paper is therefore organized as follows. Section II looks at the impact of migration
on the composition of the labor force and of income and hence on inequality, both from
a methodological (Section II-A) and an empirical point of view (Section II-B). In Section
III the focus is on the role of the size of the household and again there is first a
methodological (III-A) and then an empirical (III-B) section. Section IV analyzes, in a
similar way, changes over time in the weight of the different regions and their effect on
spatial inequality. Short concluding comments are given in Section V. Finally in three
Appendices additional results are given (Appendix A), a survey of the macroeconomic
environment in Turkey between 1960 and 2000 is presented (Appendix B) and more
precise information on the data sources are provided (Appendix C).
II. Internal Migration and Income Inequality: An Analysis via the
Decomposition of Inequality by Income Source and Population Subgroups:
A) The Methodology:
1) The Decomposition of the Gini Index by Income Sources:
Following Silber's (1989) analysis of the decomposition of income inequality, it is
possible to define the Gini Index IG of overall income inequality as:
IG = [e’] G [∑i=1 to I Sji ]
(1)
where e’ a 1 by n row vector of population shares all equal to (1/n), n being the number
of individuals in the population, I the number of income sources, Sji the n by 1 column
vector of the share of the amount of money received by individual j from income source
i in the total income of the population and G is a n by n matrix, called the G-matrix (see,
Silber 1989), whose typical element ghk is equal to 0 if h=k, to –1 if k >h and to +1 if h
>k. Note that in the vector Sji the elements are ranked by decreasing values of the total
income of the individuals. Let Xji be the vector of the shares (Sji /∑j=1 to n Sji ), ranked by
decreasing values of the total income of the individuals. The product Hi=[e’]G[Xji ] is
then called the Pseudo-Gini of income source i. However if we call Vji the vector of the
shares (Sji / ∑j=1 to n Sji ) ranked by decreasing values of the incomes received by the
individuals from source i, the product Gi = [e’] G [Vji ] is in fact the Gini index of
inequality of income source i (see, Silber, 1989, for more details).
Let S.i represent the share of income source i in the total income of the population. We
may then express (see, Silber, 1989) the Gini index IG as
IG = ∑i S.i { [ e’] G [ Sji ] } = ∑i Si Hi = ∑i Ci
(2)
where Ci represents the contribution of income source i to the overall inequality.
2) The Breakdown of the Gini Index by Population Subgroups:
Following earlier studies (see, Bhattacharya and Mahalanobis, 1967, Rao, 1969, Fei,
Ranis and Kuo, 1979, Kakwani, 1980, Lerman and Yitzhaki, 1984), Silber (1989) has
proven, using the approach based on the G-matrix which was just summarized, that
when the population is divided in subgroups, the Gini index may be decomposed into
three elements:
- a within populations contribution IW
- a between populations inequality IB
- an interaction or overlap component IO
More precisely if Pa and Wa are the shares in total population and in total income of
area a and if Ia refers to the Gini index for area a, Silber (1989) has proven that:
IW = ∑a=1 to A Pa Wa Ia
(3)
where A is the number of areas distinguished. It can also be shown that:
IB = [...Pa ...] G [...Wa ...]
(4)
where the elements in the row vector [...Pa...] and in the column vector [...Wa ...] are
ranked by decreasing average income (that is by decreasing ratios Wa /Pa ) and G is an
A by A G-matrix. Finally, the overlap component IO is defined as
IO = IG - (IW + IB )
(5)
where IG refers to the Gini index for the country as a whole.
3) Combining the Decomposition by Income Source and by Population
Subgroups:
Let the subindices a=1 and a=2 refer respectively to the urban and rural components
in the decomposition of inequality by population subgroups. One may then
decompose Ia in (3) by income source (using equations (1) and (2) ) so that the within
areas inequality IW will be written as
IW = ∑a=1 to A Pa Wa [ ∑i=1 to A Cia ]
⇔
(6)
IW = ∑i=1 to I CWi
(7)
Cwi = ∑a=1 to A Pa Wa Cia
(8)
where
refers to the contribution of income source i to the overall within income inequality
while Cia measures the contribution of source i to the overall income inequality in area
a and is defined on the basis of equation (2).
For the between areas inequality, one may proceed as follows. Let Tia refer to the total
income from source i in area a. The share Wa in (6) may then be expressed as:
Wa = ∑i=1 to I ( Tia / ∑a=1 to A Tia ) ( ∑a=1 to A Tia / ∑i=1 to I ∑a=1 to A Tia )
(9)
Combining (4) and (9) one derives, in the simple case where two areas only are
distinguished (rural versus urban areas), that
IB = [ PU PR ] G [ (∑i=1 to I TiU /T.. ) (∑i=1 to I TiR /T..)]
(10)
where the subindexes U and R refer respectively to urban and rural areas (assuming
the average income is higher in urban areas) and T.. is equal to ∑i=1 to I ∑a=1 to 2 Tia .
When the definition of the G-matrix is applied to (10) one concludes, after some
algebraic manipulations, that
IB = - PU ∑i=1 to I (TiR /T..) + PR ∑i=1 to I (TiU /T..)
(11)
IB = ∑i=1 to I Cbi
(12)
CBi = ((PR TiU - PU TiR )/T..)
(13)
where
refers to the contribution of income source i to the between areas inequality IB .
Finally recalling that Ci in (2) refers to the contribution of income source i to income
inequality for the country as a whole, one may define Coi , the contribution of income
source i to the overlap component IO in (5) as
COi = Ci - ( CWi + CBi )
(14 )
and conclude, using (2), (8), (13) and (14) that
IG = ∑i=1 to I ( CWi + CBi + COi )
(15)
B) An Empirical Illustration: Turkey in 1994
1) Basic Data on the Distribution of Incomes in Urban and Rural Areas in Turkey
in 1994:
In 1987 47.4% of the Turkish population lived in urban and 52.6% in rural areas (see,
Özmucur and Silber, 1995, mimeo). Seven years later the proportions are completely
different since 58% of the population lived then in urban and 42% in rural areas. More
details on the breakdown of the total population by area of residence (urban versus rural
areas) and population subcategory (three categories being distinguished: Wage and
Salary Earners, Daily Workers and Proprietors) are given in Table 1.
These data (lower section of Table 1) indicate that three quarters of the Wage and Salary
Earners live, as expected, in urban areas, the corresponding proportion for Daily
Workers being two thirds. For Proprietors the opposite is true since almost two thirds of
them live in rural areas. If we now look (upper section of Table 1) at the composition of
the population within urban or rural areas, it appears that Wage and Salary Earners
represent 56% of the population living in urban areas, the corresponding share being
only 24% in rural areas. The same contrast exists for Daily Workers who represent 66%
of the urban population but only 34% of the rural population. On the contrary the shares
of Proprietors in the total population are 63% in rural but only 37% in urban areas.
Table 1: Shares in Total Population of Various Categories
Category
Urban Areas
Wage and Salary 0.563
Rural Areas
Whole Country
0.241
Earners
Daily Workers
0.164
0.116
Proprietors
0.277
0.643
Together
1.000
1.000
Wage and Salary 0.763
0.237
1.000
Earners
Daily Workers
0.662
0.338
1.000
Proprietors
0.369
0.631
1.000
Together
0.580
0.420
1.000
Important differences between urban and rural areas exist also when one compares the
averages incomes of the population categories which have been distinguished. These
data are given in Table 2. Here an additional distinction has been introduced in so far as
information is also available on the type of income (Income from primary job, from
secondary job or from other sources) earned by the individuals.
Table 2 indicates clearly that, whatever the subpopulation category considered, the
average income is higher in urban than in rural areas. For Wage and Salary Earners the
difference is of 17% ((0.81-0.69)/0.69), for Daily Workers it reaches 41%. The
differences are even more striking concerning Proprietors since the latter in urban areas
earn 137% more than in rural areas. On average for all categories combined the
differential is equal to 40%.
The relative importance of the various income sources is summarized in Table 3.There
appears to be significant differences between the various categories. In urban areas
income received from the primary job represents on average 78% of the total income
while the corresponding figure in rural areas is 85%. On the contrary income from other
sources represents 18% of total income in urban areas but only 8% in rural areas. If we
now look at specific categories one observes that the weight of income from primary job
is highest among daily workers (91%) in urban areas and lowest among Proprietors in
urban areas. Note that income from secondary job represents almost 10% of total income
among Wage and Salary Earners and 7.5% among daily workers in rural areas. Income
from secondary job is much less important in urban areas. Finally other income sources
play an important role mainly among Proprietors in urban areas (it corresponds to almost
24% of their total income) and eventually also among Wage and Salary Earners in urban
areas (a share of 12.7%).
These differences in average income, together with those observed in Table 1
concerning the relative importance in the total population of the various
subpopulations, have evidently an impact on the share in total income of the various
categories. Table 4 summarizes the results.
If one first looks at the relative share of the different population categories in total
income the differences between urban and rural areas is striking. It appears that in
urban areas 54% of the total income is received by Proprietors while the share of
Wage and Salary Earners is only equal to 40%. In rural areas the share of Proprietors
reaches 75% while that of Wage and Salary Earners is only equal to 21%. Note the
small share of Daily Workers both in urban (6%) and rural (4%) areas.
If we now analyze, for each category, the relative importance of urban and rural areas
(see the lower part of Table 4) one observes that 79% of the income received by wage
and Salary Earners is earned in urban areas, the corresponding share for Daily
Workers being 73%. On the contrary only 58% of the income received by proprietors
originates in urban areas.
Table 2: Relative* Income by Income Source and Population Subgroup
Population
Category
URBAN
AREAS
Wage
and
Salary
Earners
Daily
Workers
Proprietors
Together
RURAL
AREAS
Wage
and
Salary
earners
All
income Income from Other Income
Primary and Sources
sources
Secondary
combined
Job
0.81
0.83
0.69
0.42
0.46
0.19
2.25
1.14
2.02
1.09
3.58
1.40
0.69
0.74
0.43
Daily
Workers
Proprietors
Together
URBAN and
RURAL
AREAS
combined
Wage
and
Salary
Earners
Daily
Workers
Proprietors
0.30
0.32
0.14
0.95
0.81
1.02
0.87
0.52
0.45
0.78
0.81
0.63
0.38
0.41
0.18
1.43
1.39
1.65
Together
1.00
1.00
1.00
* “Relative” means “relative to national average”, for a given income source, as
indicated in the table itself (the last row is equal to 1).
Table 3: Income Shares by Income Sources and Population Category
Population
Category
Income from Income from Income from Total
Other Sources
Primary Job
Secondary
Job
URBAN
AREAS
Wage
and 0.835
Salary
Earners
Daily
0.911
Workers
Proprietors
Together
RURAL
AREAS
Wage
and
Salary
Earners
Daily
Workers
Proprietors
Together
URBAN and
RURAL
AREAS
together
Wage
and
Salary
Earners
Daily
Workers
Proprietors
Together
0.038
0.127
1.000
0.020
0.069
1.000
0.731
0.783
0.033
0.035
0.236
0.182
1.000
1.000
0.809
0.099
0.091
1.000
0.854
0.075
0.071
1.000
0.857
0.847
0.06
0.070
0.081
0.083
1.000
1.000
0.830
0.051
0.119
1.000
0.896
0.035
0.070
1.000
0.784
0.805
0.045
0.047
0.171
0.148
1.000
1.000
Table 4: Share in Total Income of Various Categories
Category
Urban Areas
Wage and Salary 0.401
Rural Areas
Whole Country
0.206
Earners
Daily Workers
0.060
0.042
Proprietors
0.539
0.751
Together
1.000
1.000
Wage and Salary 0.790
0.210
1.000
Earners
Daily Workers
0.734
0.266
1.000
Proprietors
0.581
0.419
1.000
Together
0.659
0.341
1.000
So far we have only taken a look at the average income earned by the different
population categories. A more complete view of income distribution in Turkey requires
evidently that one analyzes also the distribution of incomes separately for each
population category. Such an issue becomes even more important once one recalls the
observation made earlier concerning the intensity of the migration flows (from rural to
urban areas) which have taken place in Turkey during the period 1987-1994. To measure
the degree of inequality of the income distributions considered we have used the Gini
Index and applied the methodology presented in Section II-A.
2) Decomposing the overall income inequality in Turkey in 1994:
The results of this breakdown are presented in Tables 5 to 7.
Table 5 gives the contribution of each income source to the Gini index of each
population subcategory. For both urban and rural areas and for each category the
greatest contribution to the Gini index is, as expected, that of the primary income. Its
relative contribution is however generally higher in rural than in urban areas. For
example for Proprietors, the most important category in rural areas, the contribution
of primary income to the Gini index of total income among proprietors in rural areas
is equal to 88% while for Wage and Salary Earners, the most important category in
urban areas, 75% of the Gini index of total income is explained by primary income.
In Table 6 the contribution of each income source to the three components of
inequality (between groups, within groups and overlap) is given separately for urban
and rural areas.
16
Table 5: Decomposition of Within Groups Inequality by Income Source For
Each Population Subcategory
Population
Subcategory
URBAN
AREAS
Wage
and
Salary
Earners
Daily
Workers
Proprietors
All categories
combined
RURAL
AREAS
Wage
and
Salary
Earners
Daily
Workers
Proprietors
All categories
combined
Total
Inequality
(Gini Index)
Income from Income from Other Income
Sources
Primary Job
Secondary
Job
0.102
0.077
0.007
0.018
0.004
0.003
0.000
0.001
0.088
0.194
0.059
0.139
0.003
0.010
0.026
0.045
0.020
0.014
0.003
0.003
0.002
0.002
0.000
0.000
0.217
0.239
0.192
0.208
0.013
0.016
0.012
0.015
17
Concerning urban areas one may first note that if the highest contribution to both the
between and the within groups inequality is that of primary income, this income
source plays, both in absolute and relative terms, a greater role for the between groups
inequality (84% of the total between categories inequality) than for the within groups
inequality (72% ). In rural areas the situation is reversed. There the contribution of
primary income is higher, both in absolute and relative terms, for the within than for
the between groups inequality. In the former case its contribution represents 87% of
the inequality, in the latter case 76%. The other income sources play a relatively
minor role, except in the case of within groups inequality in urban areas since in that
case their share in inequality is equal to 23% (.045/.194).
Table 7 summarizes all the previous results since it presents the breakdown of the
overall inequality in Turkey in 1994 by type of inequality, by area of residence and by
income source
It appears that inequality within categories (the latter referring to the three types of
workers: Wage and Salary Earners, Daily Workers and Proprietors) is much more
important, for Turkey as a whole, than the between categories inequality and that 77%
(.223/.289) of this within groups inequality is contributed by urban areas. Remember
that the share of these areas in the total population is only 58% (see, Table 2) whereas
that in total income is only 66% (see, Table 4). One may also observe that the overlap
component (.178) for Turkey as a whole is much more important than the between
groups inequality, the latter referring here to urban versus rural areas, all categories
(that is all types of workers) combined. In other words there is an important dispersion
of income within urban and within rural areas so that the two distributions largely
overlap. The results concerning the within groups inequality are very similar to those
presented in table 6, the only difference being that the data of Table 6 have been
18
multiplied by the product of the shares of urban (rural) areas in the total population
and in total income.1
1
e.g. .223, on the first line of table 10, is simply the product of .583 which appears on the first line of
Table 9, by .58, the share of urban areas in the total population and by .66, the share of urban areas in
total income.
19
Table 6: Decomposition of Total Inequality by Income Source separately for
Urban and Rural Areas.
Population
Category and
Type
of
Inequality
URBAN
AREAS
Overall Gini
Index
Between
Categories
Inequality
(Gini Index)
Within
Categories
Inequality
(Gini Index)
Measure
of
Overlap
RURAL
AREAS
Overall Gini
Index
Between
Categories
Inequality
(Gini Index)
Within
Categories
Inequality
(Gini Index)
Measure
of
Overlap
Contribution
to Inequality
Income from Income from Other Income
Primary Job
Secondary
Sources
Job
0.583
0.299
0.251
0.012
0.035
0.194
0.139
0.010
0.045
0.123
0.093
0.017
0.013
0.239
0.208
0.016
0.015
0.090
0.464
0.103
20
Table 7: Decomposition of Overall Income Inequality in Turkey (1994) by Population Subcategory and Income Sources
Overall
Contribution
Inequality
Between
Urban and
Rural
Areas
0.079
Within
Urban and
Rural
Inequality
0.289
Contribution
Primary Job
of Contribution Contribution of
of Secondary Income Sources
Job
0.223
Contribution
of
Urban
Areas
Between
Workers
categories
Inequality
0.114
0.096
21
0.005
0.013
Other
Table 7 (cont.)
Within
Workers
categories
Inequality
Overlap
Contribution
of
Rural
Areas
Between
Workers
categories
Inequality
of Contribution Contribution of
of Secondary Income Sources
Job
0.004
0.17
Overall
Contribution
Contribution
Primary Job
0.074
0.053
Wage and 0.039
Salary
Earners
0.029
0.003
0.007
Daily
0.001
Workers
Proprietors 0.034
0.034
0.066
0.001
0.000
0.000
0.023
0.001
0.010
0.018
0.013
0.002
0.002
22
Other
Table 7 (end)
Within
Workers
categories
Inequality
Overlap
Overlap
between
Urban and
Rural
Areas
Total
Inequality
for Turkey
(Gini
Index)
of Contribution Contribution of
of Secondary Income Sources
Job
0.002
0.002
Overall
Contribution
Contribution
Primary Job
0.034
0.030
Wage and 0.003
Salary
Earners
Daily
0.000
Workers
Proprietors 0.031
0.015
0.178
0.002
0.0005
0.0005
0.000
0.000
0.000
0.027
0.002
0.002
0.546
23
Other
III) Internal Migration and the Impact of Changes in Household Size on Income
Inequality:
A) The Methodology:
1) A Simple Formulation for the Decomposition of Income Inequality by
Population Subgroups:
Using concepts defined previously let GTOT, GBET, GWITH and GOVERL
refer
respectively to the overall value of the Gini Index of income inequality in a given
country, to the between regions inequality in this country, to the within regions
income inequality and finally to the residual term of the decomposition of overall
inequality which measures in fact the degree of overlap between the income
distributions of the urban and rural areas. We may then write (see, Silber, 1989, and
Deustch and Silber, 1999) that
GTOT = GBET + GWITH + GOVERL
(16)
The within areas inequality GWITH may be expressed (see, expression (6) and for more
details, Silber, 1989) as
GWITH = ∑r=1 to R pr sr Gr
(17)
where pr refers to the population share of region r, sr to the share of region r in the
total income of the country, Gr to the Gini index of income inequality within region r
while R represents the total number of regions in the country. Let us call respectively
ym and ymr the average incomes in the whole country and in region r.
Since sr may be also expressed as
sr = pr (ymr/ym)
(18)
we may also write (17) as
GWITH = ∑r=1 to R (pr)2 (ymr/ym) Gr
(19)
24
We have therefore expressed the within areas inequality index GWITH as a function of
only three sets of variables: the shares of the various regions in the total population of
the country, the ratios of the average incomes in the various regions over the average
income in the whole country and the within regions Gini indices.
We will now show that is possible to express the between areas inequality index GBET
as a function of only two sets of variables: the shares of the various regions in the total
population and the ratios of the average incomes in the various regions over the
average income in the whole country.
It may in fact be shown (see, Silber, 1989) that if [pr ] represents the row vector of the
population shares of the various regions while [sr ]’ refers to the column vector of the
shares of the various regions in the total income of the country, the regions in both
vectors being ranked by decreasing values of the regional average incomes, the
between areas Gini inequality index GBET may be expressed (cf., expression (10) in
Section II-A-3) as:
GBET = [ pr] G [ sr]’
(20)
where G is the G-matrix which was defined previously.
Combining then (18) and (20) one derives that
GBET = [ pr] G [ pr (yr / ym )]’
(21)
where as before the elements of the row vector [pr ] and of the column vector
[ pr (yr / ym )]’ are ranked by decreasing values of the regional incomes yr .
2) Measuring the Respective Impacts of Income and Size of the Household on
Inequality Differences:
In the previous section no attention was given to two important questions:
25
-
on which measure of the welfare of the household members should the
inequality analysis or comparison be based?
-
which are the units (households, individuals) whose distribution of welfare we
want to analyze?
These issues have been analyzed in Danziger and Taussig (197?) and we summarize
here the main ideas.
a) Measuring the welfare of household members:
The main question here is in fact to determine which part of the goods and services
consumed by the household should be considered as purely private goods (whose
consumption cannot be shared, e.g. food) and which part as public goods (e.g. a
refrigerator). Buhman et al. (197?) proposed a nice formulation to tackle this problem
by expressing the welfare xi of household members as
xi = yi / (ni )a
(22)
where yi is the total income of the household, ni is the size of the household and a is a
parameter included in the interval [0,1]. It may be observed that if a = 0, the welfare
of the household members is equal to the total household income, so that it is then
assumed that all goods and services are considered as public goods. On the contrary
when a = 1, the welfare indicator is equal to the per capita income, in which case one
supposes that all goods and services are private. A more general case occurs when
0<a<1 which implies that part of the goods and services consumed are public, part
private.
It is also possible to make a difference between adults and children and write (cf.,
Coulter et al., (19??) ) that
ni = ai + λ ci
(23)
26
where ai is the number of adults in the household and ci is the number of children,
while λ is a parameter indicating how the consumption of a children should be
converted into an adult’s consumption.
b) The selection of the unit whose welfare distribution is analyzed:
This question is different from the previous one. Whatever measure of welfare is
analyzed, we may ask whether we want to look at the inequality between households
or between individuals. Given that four measures of welfare and two units of
observation (individuals or households), we will hence have eight different ways of
measuring inequality.
To measure inequality we use the algorithm based on the G-matrix (see expressions
(1) or (4) ) and express the Gini index of inequality Gr within a given region r as
Gr = [e’] G [s]
(24)
where e’ is a row vector of the shares in the total population of the different subgroups
distinguished (these subgroups may for example be the deciles of the total population
in region r), G the G-matrix defined earlier and s a column vector of the shares of the
different subgroups in total income, the elements of e’ and s being ranked by
decreasing values of the average income of each subgroup.
Calling respectively hi and ni the number of households and individuals in subgroup i
and yi the average household income in this subgroup, we summarize in Table 8 the
way expression (24) will be expressed in the eight cases distinguished.
c) Analyzing the respective impacts of household income and size on inequality
comparisons:
Let us for example compare the value of the between households Gini index of
income per equivalent adult in two regions. Table 8 indicates that for region r this
index may be expressed as
27
Gr = [hi /∑ihi]’G[(yi /(ni)0.5)hi /∑i[(yi/(ni)0.5)hi ]
(25)
Assuming that the data are given by decile so that (hi /∑ihi) = 0.1 for every i, we may
write that Gr as a function f (yir,nir ) where the subscript r indicates to which region the
function r refers. The difference ∆G between the values of the Gini index in two
regions r and s may therefore be expressed as
∆G = f (yir,nir ) - f (yis,nis ) = 0.5(C1 + C2 ) + 0.5(C3 + C4 )
(26)
where
C1 = f (yir,nir ) - f (yir,nis )
(27)
C2 = f (yis,nir ) - f (yis,nis )
(28)
C3 = f (yir,nir ) - f (yis,nir )
(29)
C4 = f (yir,nis ) - f (yis,nis )
(30)
It is easy to show that the expressions 0.5(C1 + C2 ) and 0.5(C3 + C4 ) measure
respectively the impact of differences between regions r and s in the sizes of the
households and in total
28
Table 8: Various ways of expressing the Gini index.
Welfare
Inequality between households
Inequality between individuals
[hi /∑i hi ]’ G [yi /∑I yi ]
[ni /∑i ni ]’ G [yi /∑i yi ]
Per capita
[hi /∑i hi ]’ G [(yi /ni)hi /∑i (yi
[ni /∑i ni ]’ G [(yi /ni)ni /∑i (yi /ni)ni]
income
/ni)hi]
Income per
[hi /∑ihi]’G[(yi /(ni)0.5)hi
[ni /∑ini]’G[(yi /(ni)0.5)ni
equivalent
/∑i[(yi/(ni)0.5)hi ]
/∑i[(yi/(ni)0.5)ni ]
[hi/∑ihi]’G[(yi/(ai+0.5ci)0.5)hi
[ni/∑ini]’G[(yi/(ai+0.5ci)0.5)ni
Indicator
Total
household
income
adult
Alternative
formulation /∑i[(yi /(ai+0.5ci)0.5)hi]
/∑i[(yi /(ai+0.5ci)0.5)ni]
of the
income per
equivalent
adult
29
household income. A similar decomposition may be obtained when the alternative
definition of income per equivalent adult is used.
In the case where one measures inequality between individuals expression (24) will be
expressed as
Gr = [ni /∑ini]’G[(Yi /(ni)0.5)ni /∑i[(Yi/(ni)0.5)ni ]
(31)
where Yi is the total income of all households belonging to subgroup i (e.g. deciles of
households).
The ratio (Yi /(ni)0.5) may be also expressed as
(Yi /(ni)0.5) = (Yi /hi) / ((ni)0.5/ hi) = (zi / ei )
(32)
where zi and ei measure respectively the per household income and the number of
equivalent adults per household.
The difference ∆G between the values of the Gini index in two regions r and s may
therefore be expressed in this case as
∆G = f (zir, eir, nir ) - f (zis, eis, nis )
(33)
Using decomposition techniques quite similar to those given in expressions (27) to
(30) but applied to the case where we have three explanatory variables (zi, ei, and ni ),
it is possible to measure the respective impacts on this gap ∆G of differences between
the subgroups (e.g. household deciles) in the total number of individuals (role of
differences in ni ), in the total household incomes (role of zi) and in the size of the
households (differences in ei).
B) An Empirical Illustration: Spatial Inequality in Turkey in 1994:
1) Differences between urban and rural areas:
30
The results of this analysis are reported in Tables 9 to 12. Table 9 indicates that the
average income, whether of households or of individuals, in urban areas is about 25%
higher that in Turkey as a whole while in rural areas it is about 70% lower. Table 10
then indicates that, whatever concept of inequality one uses, inequality is much higher
in urban than in rural areas. If one decomposes the overall inequality in Turkey into
between areas (urban and rural), within areas and an overlapping component, Table
11 shows that close to 50% of the overall inequality2 is attributed to within areas
inequality, the between areas inequality accounting for 34 to 37% and the overlapping
component for 16 to 18% of the overall inequality.
Note also (see, Table 10) that the difference between the two areas in within areas
inequality is highest when per capita income is the measure of welfare chosen and
lowest when total household income is used, this being true for both inequality
between households and between individuals.
In Table 12 we have decomposed the difference between the inequality of per capita
or per equivalent adult income in urban and rural areas into two components (see, the
methodology exposed in section III-A-2-c) measuring respectively differences
between the two areas in total household income and in the average size of the
households. It appears that the latter component (role of he household size) explains
approximately 60% of the difference while the former component (role of total
household income) accounts for the remaining 40%.
The important role played by differences in the size of the household appears clearly
in Table 9 which shows that not only the average size of the household is smaller in
urban areas but also its standard deviation and th coefficient of variation of the size of
the households. Moreover the data indicate clearly that the average size of the
2
Note that the overall inequality levels in this section are smaller than those found in Section II
because we use quintiles while in Section II more detailed data were available (by population
31
household, in both urban and rural areas increases with the average income, whether
of households or of individuals.
Remembering that our analysis is based on date collected for quintiles, we have
computed also the ratio of the average size of the households in the richest over that in
the poorest quintile. This ratio turns out to be equal to 1.12 in urban areas and to 1.56
in rural areas. We have similarly computed the ratio of the average total household
income in the richest over that in the poorest quintile and found that this ratio was
equal to 11.8 in urban and 8.6 in rural areas. Combining these two types of results on
derives that the ratio of the per capita income in the richest over that in the poorest
quintile is equl to 10.3 in urban and 5.5 in rural areas.
What these data imply is that migration from rural to urban areas induces an increase
in the inequality of per capita income (whether one talks about inequality between
households or between individuals) not only because the inequality of total household
incomes is higher in urban areas but probably also because that of household size is
higher in rural areas.
2) Regional Differences:
The results of this analysis are presented in Tables 13 to 16. Table 13 indicates that,
whether for household or per capita income, the average income is highest in
Marmara and the region of the Aegean Sea and lowest in Eastern and Souther
Anatolia (see, in Table 13, the data on the relative income of the various regions, the
comparison being made with Turkey as a whole). Table 14 then shows that the within
region inequality is highest for total household income and lowest for per capita
income, this being true whether one looks at the between households or the between
individuals inequality.The data of table 14 indicate also clearly, that, whatever
subcategories, income source, etc…).
32
concepts of inequality are used, the within region inequality is highest in the three
richest regions (Marmara, Aegean Sea and the Mediterranean region) and lowest in
the two poorest regions (Eastern and Southern Anatolia). Note also that the greatest
difference between the regions is observed when per capita income is the measure of
welfare used and the smalles one when total household income is used, this result
being similar to the one we observed when comparing urban and rural areas.
A look at Table 15 shows that when regions are compared the between regions
inequality is much higher than the within regions inequality, whatever measure of
welfare is used and whether one looks at inequality between households or
individuals. In fact 40 to 60% of the overall inequality is explained by between
regions and only 15 to 18% by within regions differences.
In table 16, we have decomposed the difference between inequality within a given
region and that in Turkey as a whole into three components (see, the methodology
presented in section III-A-2-c). The first component is due to differences between the
quintiles in population shares (this is a consequence of the fact that whereas the share
of the households in each quintile is by definition equal to 20%, that of individuals is
not necessariy equal to 20%). The second component reflects differences between the
quintiles in total household income while the third element is a consequence of
differences in the size of the households. It appears that in the poor regions (Eastern
and Southern Anatolia) the component reflecting differences in total household
income accounts for 80 to 90% of the total gap while that due to differences in
household size explains 10 to 20%. In rich areas (Marmara and Aegean Sea) the
results are not clear, total household income contributing to even more than 100% of
the overall difference in Marmara but to only 60$ in the Aegean Sea area.
33
As far as the size of the households is concerned it appears (see, Table 12) that first
this average size is much higher in poor regions (the average being there equal to 5.5
to 5.7 individuals) than in rich regions (average varying from 3.85 to 4.12). Similarly
the standard deviation of household size is lower in rich than in poor regions but the
results for the coefficient of variation of household size are not clear.
When comparing household size with relative (to Turkey as a whole) total household
income or per capita income in the various regions we find a negative correlation
between average household size and total household income (correlation: -0.61) or per
capita income (correlation: -0.77). A similar analysis shows a negative correlation
between the standard deviation of household size and relative total household income
(correlation:-0.48) or per capita income (correlation:-0.55).
The results of this regional analysis are thus quite similar to those derived in the urban
versus rural areas comparison, though sometimes less clear-cut conclusions may be
derived because some of the regions are not specifically urban or rural areas. But as a
whole they seem to confirm the important role played by internal migration which
affects inequality both through its impact on the inequality of total household incomes
and on the size of the households.
34
Table 9: Summary Data for the Urban and Rural Areas
Region
Share in
Total
Number of
Households
Share in
Total
Number of
Individuals
Average
Share
Size of
in
Household
Total
Income
Standard
Deviation
of Size of
Household
Urban Areas
Rural Areas
0.562
0.438
0.536
0.464
0.689
0.311
0.22
0.68
4.24
4.71
Coefficient
of
Variation
of Size of
Household
0.052
0.144
Relative
Relative
Income of Income of
Household Individuals
1.23
0.71
1.29
0.67
Table 10: Summary Data on Within Areas Inequality
Region
Urban Areas
Rural Areas
Turkey as a
whole
Between
Between
Between
Between
Between
Between
Between
Between
Households- Households- Households- Households- Individuals- Individuals- Individuals- IndividualsAlternative
Per
Per Capita
Per
Alternative
Per
Per Capita
Per
Per
Equivalent
Income
Household
Per
Equivalent
Income
Household
Equivalent
Adult
Income
Equivalent
Adult
Income
Income
Income
Income
Income
0.458
0.443
0.451
0.447
0.452
0.439
0.446
0.442
0.384
0.321
0.353
0.351
0.370
0.313
0.342
0.341
0.442
0.414
0.428
0.425
0.434
0.408
0.421
0.418
35
Table 11: Summary Table of Decomposition of Inequality for Urban and Rural Areas
Concept of
Inequality and
Measure of
Welfare of
Household
INEQUALITY
BETWEEN
HOUSEHOLDS
Inequality of
Household
Income
Inequality of
Per Capita
Income
Inequality of
Per Equivalent
Adult Income
Inequality of
Alternative
Measure of Per
Equivalent
Adult Income
Share of
Share of
Overlap Share of
Within
Between
Total
Overlap
Within
Between
Areas
Areas
(Whole
in Total
Areas
Areas
Country) Inequality Inequality
Inequality Inequality Inequality
Inequality
in Total
in Total
Inequality Inequality
0.479
0.178
0.229
0.072
0.37
0.48
0.15
0.443
0.149
0.216
0.078
0.34
0.49
0.18
0.462
0.164
0.223
0.075
0.35
0.48
0.16
0.458
0.163
0.221
0.074
0.36
0.48
0.16
36
Concept of
Inequality and
Measure of
Welfare of
Household
INEQUALITY
BETWEEN
INDIVIDUALS
Inequality of
Household
Income
Inequality of
Per Capita
Income
Inequality of
Per Equivalent
Adult Income
Inequality of
Alternative
Measure of Per
Equivalent
Adult Income
Share of
Share of
Overlap Share of
Within
Between
Total
Overlap
Within
Between
Areas
Areas
(Whole
in Total
Areas
Areas
Country) Inequality Inequality
Inequality Inequality Inequality
Inequality
in Total
in Total
Inequality Inequality
0.467
0.183
0.220
0.064
0.39
0.47
0.14
0.433
0.153
0.207
0.073
0.35
0.48
0.17
0.450
0.168
0.214
0.068
0.37
0.48
0.15
0.446
0.167
0.212
0.067
0.37
0.48
0.15
37
Table 12: Decomposition of the Difference* in Inequality between Urban and Rural Areas
Inequality Inequality Inequality Inequality Inequality Inequality
Between
Between
Between
Between
Between
Between
Households Households Households Individuals Individuals Individuals
Gap due to Gap due to
Total
Gap due to Gap due to
Total
Difference difference differences Difference difference difference
in
in Shares
between
in size of
in
between
household
Urban and household households Urban and
incomes
Rural
incomes
Rural
Areas
Areas
+0.122
+0.076
+0.046
+0.126
+0.004
+0.078
Comparison of
(0.097)
(+0.075)
(+0.022)
(+0.104)
(+0.006)
(+0.077)
Urban and
Rural Areas
Areas
•
Inequality
Between
Individuals
Gap due to
differences
in size of
households
+0.044
(+0.021)
The number not in parenthesis refer to the case where the welfare of the household (individual) is measured by the per capita income while that in parenthesis
corresponds to the case where this welfare is assumed to be equal to the per equivalent adult income. The component “population shares “ which appears in the case
where one measures inequality between individuals is due to the fact that the data were given for deciles of the household population, not of the population of
individuals.
38
Table 13: Summary Data for the Regions in 1994
Relative
Relative
Income of Income of
Household Individuals
0.36
0.52
0.31
0.44
Coefficient
of
Variation
of Size of
Household
0.088
0.136
0.068
0.102
1.45
0.89
0.88
0.86
1.56
1.02
0.87
0.90
4.68
5.51
0.43
0.75
0.093
0.136
0.85
0.80
0.81
0.65
5.72
0.56
0.098
0.61
0.47
Region
Share in
Total
Number of
Households
Share in
Total
Number of
Individuals
Average
Share
Size of
in
Household
Total
Income
Standard
Deviation
of Size of
Household
Marmara
Aegean
Mediterranean
Central
Anatolia
Black Sea
Eastern
Anatolia
Southern
Anatolia
.266
.157
.125
.179
.247
.136
.127
.172
.386
.139
.111
.154
4.12
3.85
4.54
4.27
.128
.071
.135
.088
.109
.057
.074
.096
.045
39
Table 14: Summary Data on Within Regions Inequality in 1994
Between
Between
Between
Between
Between
Between
Between
Between
Households- Households- Households- Households- Individuals- Individuals- Individuals- IndividualsAlternative
Per
Per Capita
Per
Alternative
Per
Per Capita
Per
Per
Equivalent
Income
Household
Per
Equivalent
Income
Household
Equivalent
Adult
Income
Equivalent
Adult
Income
Income
Income
Income
Income
0.490
0.460
0.475
0.472
0.482
0.454
0.468
0.466
Marmara
0.401
0.347
0.375
0.373
0.388
0.340
0.365
0.363
Aegean
0.423
0.399
0.411
0.407
0.417
0.395
0.406
0.402
Mediterranean
0.412
0.380
0.397
0.394
0.400
0.372
0.386
0.384
Central
Anatolia
0.414
0.376
0.395
0.394
0.405
0.370
0.388
0.387
Black Sea
0.342
0.279
0.311
0.309
0.329
0.271
0.301
0.299
Eastern
Anatolia
0.351
0.306
0.329
0.325
0.345
0.304
0.325
0.321
Southern
Anatolia
0.442
0.414
0.428
0.425
0.434
0.408
0.421
0.418
Turkey as a
whole
Region
40
Table 15: Summary Table of Decomposition of Inequality
Concept of
Inequality and
Measure of
Welfare of
Household
INEQUALITY
BETWEEN
HOUSEHOLDS
Inequality of
Household
Income
Inequality of Per
Capita Income
Inequality of Per
Equivalent Adult
Income
Inequality of
Alternative
Measure of Per
Equivalent Adult
Income
Total
(Whole
Country)
Inequality
Within
Between
Regions
Regions
Inequality Inequality
Overlap
Share
Share of
Share of
of
Within
Between
Overlap
Regions
Regions
Inequality Inequality in Total
in Total Inequality
in Total
Inequality Inequality
0.549
0.339
0.085
0.125
0.62
0.15
0.23
0.442
0.182
0.0078
0.182
0.41
0.18
0.41
0.494
0.261
0.081
0.152
0.53
0.16
0.31
0.490
0.259
0.081
0.150
0.53
0.17
0.31
41
Concept of
Inequality and
Measure of
Welfare of
Household
INEQUALITY
BETWEEN
INDIVIDUALS
Inequality of
Household
Income
Inequality of Per
Capita Income
Inequality of Per
Equivalent Adult
Income
Inequality of
Alternative
Measure of Per
Equivalent Adult
Income
Total
(Whole
Country)
Inequality
Within
Between
Regions
Regions
Inequality Inequality
Overlap
Share
Share of
Share of
of
Within
Between
Overlap
Regions
Regions
Inequality Inequality in Total
in Total Inequality
in Total
Inequality Inequality
0.548
0.352
0.079
0.117
0.64
0.14
0.21
0.439
0.191
0.073
0.175
0.44
0.17
0.40
0.493
0.273
0.076
0.144
0.55
0.15
0.29
0.488
0.269
0.076
0.143
0.55
0.16
0.29
42
Table 16: Decomposition of the Difference* in Inequality between Turkey as a Whole and the Various Regions
Inequality Inequality Inequality Inequality Inequality Inequality Inequality
Between
Between
Between
Between
Between
Between
Between
Households Households Households Individuals Individuals Individuals Individuals
Gap due to Gap due to Gap due to
Total
Gap due to Gap due to
Total
Difference difference differences Difference difference difference differences
in size of
in
in Shares
between
in size of
in
between
household households
Turkey
household households
Turkey
incomes
and the
incomes
and the
Region
Region
-0.046
-0.050
+0.004
-0.046
0
-0.051
+0.004
Marmara
(0.047)
(-0.049)
(+0.002)
(-0.048)
(0)
(-0.050)
(+0.002)
+0.067
+0.042
+0.024
+0.068
+0.003
+0.043
+0.022
Aegean
(+0.053)
(+0.041)
(+0.011)
(+0.057)
(+0.004)
(+0.042)
(+0.011)
+0.015
+0.019
-0.004
+0.013
-0.001
+0.019
-0.004
Mediterranean
(-0.002)
(-0.002)
(-0.002)
(+0.017)
(+.019)
(+0.015)
(+0.018)
+0.034
+0.030
+0.004
+0.037
+0.002
+.0032
+0.003
Central
(+0.031)
(+0.030)
(+0.002)
(+0.035)
(+0.002)
(+0.031)
(+0.001)
Anatolia
+0.033
+0.029
+0.005
+0.038
0
+0.029
+0.009
Black Sea
(+0.033)
(+0.029)
(+0.004)
(+0.033)
(0)
(+0.029)
(+0.004)
+0.135
+0.104
+0.031
+0.137
+0.002
+0.105
+0.030
Eastern
(+0.117)
(+0.l02)
(+0.015)
(+0.120)
(+0.003)
(+0.103)
(+0.014)
Anatolia
+0.108
+0.097
+0.010
+0.104
0
+0.094
+0.010
Southern
(+0.099)
(+0.094)
(+0.005)
(+0.096)
(0)
(+0.091)
(+0.005)
Anatolia
Region
•
The number not in parenthesis refer to the case where the welfare of the household (individual) is measured by the per capita income while that in parenthesis
corresponds to the case where this welfare is assumed to be equal to the per equivalent adult income. The component “population shares “ which appears in the case
where one measures inequality between individuals is due to the fact that the data were given for deciles of the household population, not of the population of
43
IV) Analyzing the changes over time in regional inequality:
A) Decomposing the Change Over Time in the Gini Index:
Let the subscripts 0 and 1 refer respectively to times 0 and 1. Using (4) it may easily
be shown3 that the change ∆GWITH in the within areas inequality index GWITH may be
expressed as
∆GWITH = A + B
(34)
where
A = ∑r=1 to R {[(1/2)((pr0)2 + (pr1)2 )][( (ymr1/ym1) Gr1 ) - ( (ymr0/ym0) Gr0 )]}
(35)
and
B = ∑r=1 to R {[(1/2)(((ymr1/ym1) Gr1 ) + ( (ymr0/ym0) Gr0 ))] [((pr1)2 - (pr0)2 )]}
(36)
However expression A in (35) may be also written4 as
A=C+D
(37)
where
C = ∑r=1 to R {[(1/4)((pr0)2 + (pr1)2 )][((ymr1/ym1) + (ymr0/ym0)) (Gr1 - Gr0 )]}
(38)
D = ∑r=1 to R {[(1/4)((pr0)2 + (pr1)2 )][(Gr1 + Gr0 ) ((ymr1/ym1) - (ymr0/ym0))]}
(39)
Combining now expressions (34) to (39) we conclude that
∆GWITH = B + C + D
(40)
where B, C and D measure respectively the impacts of the change in the population
shares of the various regions, in the within regions inequality and in the relative (to
the overall average income) income of the different regions.
Using (21) we may now similarly decompose the change ∆GBET in the between areas
inequality. This change will be expressed as
∆GBET = [ pr1] G [ pr1 (yr1 / ym1 )]’ - [ pr0] G [ pr0 (yr0 / ym0 )]’
3
(41)
We use here again the well known result according to which (ab-cd) = ((a+c)/2) (b-d) + ((b+d)/2)(a-c)
Alternative decompositions may be derived and one could think of using an average of all the
possible decompositions.
4
44
It may then be shown that this change may be expressed as
∆GBET = E + F
(42)
where
E = (1/2) (H + I)
(43)
and
F = (1/2) (J + K)
(44)
with
H = { [ pr1] G [ pr1 (yr1 / ym1 )]’} - {[pr0] G [ pr0 (yr1 / ym1 )]’}
(45)
I = { [ pr1] G [ pr1 (yr0 / ym0 )]’} - {[pr0] G [ pr0 (yr0 / ym0 )]’}
(46)
J = { [ pr1] G [ pr1 (yr1 / ym1 )]’} - {[pr1] G [ pr1 (yr0 / ym0 )]’}
(47)
K = { [ pr0] G [ pr0 (yr1 / ym1 )]’} - {[pr0] G [ pr0 (yr0 / ym0 )]’}
(48)
It may be observed that F measures the impact on the change in the between areas
inequality of changes in the relative (to the overall average income) income of the
various regions while E measures the effect of the change in the shares of the different
regions in the total population.
Combining finally (16) with expressions (40) and (48) we conclude that the change
∆G in the overall inequality may be expressed as
∆G = (B + E) + C + (D + F) + ∆GOVERL
(49)
where (B + E) , C, (D + F) and ∆GOVERL measure respectively the impacts of the
change in the population shares of the various regions, in the within regions income
inequality, in the relative income of the regions and in the amount of overlapping
between the distributions of income in the different regions.
45
B) An empirical illustration: Changes Over Time in Turkish Regional Inequality
In order to be able to compare changes in overall inequality over time we have
grouped the regions in five entities: Marmara and Aegean area, Mediterranean
area, Central Anatolia, Black Sea, Eastern and Southern Anatolia. In table A-1 to
A-4 we present data on the regional distribution of income in 1968, 1973, 1987
and 1994. As may be observed the classification in regions varies from one year to
the other but in grouping the regions in five big entities we manage to take a look
at the changes that took place over time.
Such a grouping implies evidently that the number of observations on which the
comparison is based may be different from one year to the other. Assume for
example that the data on the distribution of incomes are given at the level of
quintiles in 1987 and 1994 but that the regions of Marmara and of the Aegean Sea
were grouped in one area in 1987 but not in 1994. We may evidently group these
two regions in one area in 1994 but then the computation of the inequality index
for this aggregate area in 1994 will be based on 10 and not 5 observations. The
comparison of the inequality in 1994 and 1987 is hence not rigorous since the
grouping of data is not identical in both years. Another issue is that the income
surveys of the different years are not always perfectly comparable, given the
definitions of the units of observations or of income that were used. The only
cases where the data are apparently compatible are 1987 and 1994. In what
follows we will ignore these issues and decompose the changes in inequality in
each of the three sub-periods distinguished (1968-1973, 1973-1987 and 19871994).
In Table 17 we give the share of each of the five aggregate areas in the total
Turkish population in 1968, 1973, 1987 and 1994. It appears that the share of the
46
Marmara and Aegean areas increased considerably during these 26 years since it was
equal to 30.7% in 1968 and 42.3% in 1994. On the contrary the shares of three other
areas decreased during the same period: for the Mediterranean area from 15.3% to
12.5%, for Central Anatolia from 22.6% to 17.9% and for the Black Sea area from
17.7% to 12.8%). The share of Eastern and Southeastern Anatolia on the contrary
barely varied over time.
Assuming that the income data for the four years may be compared, we may observe
in Table 18 that for some areas the changes in income shares were even more
important (Marmara and Aegean areas: 0.393 to 0.525; Central Anatolia: from 0.231
to 0.154 and Black Sea: 0.147 to 0.109) while for the two other areas (Mediterranean
Sea and Eastern and Southern Anatolia) there were no important changes.
Data on within area inequality are presented in Table 19. Assuming again that the data
are comparable we observe as a whole a decrease in inequality in each region between
1968 and 1987 and for four of the five regions an increase in inequality between 1987
and 1994, an increase being observed during this period only in Eastern and
Southeastern Anatolia.
Using the methodology exposed in section IV-A we present in tables 20 and 21 the
breakdown of changes over time in the between and within areas inequality into
various components. For the variation in between inequality there are two
components, one reflecting variations in the population shares of the various regions
and one measuring the impact of changes in the relative income (relative to the mean
for Turkey as a whole) of the various regions. The results refer to total household
income and we concentrate our analysis on the 1987-1994 period. It appears that the
between regions inequality increased from 0.091 to 0.113 and that almost the whole
47
change (+0.022) was a consequence of changes in regional population shares
(+0.021).
For the analysis of changes in within regions inequality the theoretical breakdown is
different. Since the within regions inequality is a weighted sum of the inequality
observed for each region, the weights being equal to the product of the population and
income shares of each region, the change in the overall within regions inequality will
include three elements reflecting respectively variations in population shares, relative
incomes and within region inequality. The results of the analysis are presented in
Table 21. The overall within regions inequality increased from 0.0419 to 0.0435. Here
again the greatest part of the change is a consequence of variations in regional
population shares (+0.0020 out of a change of +0.0016). Changes in regional relative
incomes did not play any role. One may observe however the negative (but weak)
effect of changes in Gini indices, this implying that if population and relative incomes
had remained constant, overall within regions inequality would have eventually
decreased.
The central role played here also by variations in population shares confirms the
important impact that internal migration has on income inequality, whether it be
between or within regions inequality. This study of changes in inequality over time
has been applied only to the case of inequality between households of total household
income. From the results of the analysis presented in section III-B we may expect that
similar conclusions would have been derived if the analysis had concerned per capita
or per equivalent adult income or if we had looked at between individuals inequality.
48
Table 17: Population shares of the various regions over time
Region
Marmara and
Aegean area
Mediterranean area
Central
Anatolia
Black Sea
Eastern and
Southeastern
Anatolia
Population
share in 1968
0.307
Population
share in 1973
0.337
Population
share in 1987
0.370
Population
share in 1994
0.423
0.153
0.152
0.134
0.125
0.226
0.219
0.243
0.179
0.177
0.138
0.149
0.147
0.106
0.147
0.128
0.145
49
Table 18: Income shares of the various regions over time
Region
Marmara and
Aegean area
Mediterranean area
Central
Anatolia
Black Sea
Eastern and
Southeastern
Anatolia
Income share
in 1968
0.393
Income share
in 1973
0.377
Income share
in 1987
0.450
Income share
in 1994
0.525
0.114
0.132
0.107
0.111
0.231
0.234
0.215
0.154
0.147
0.115
0.158
0.099
0.089
0.139
0.109
0.102
50
Table 19: Gini index of the various regions over time
Region
Marmara and
Aegean area
Mediterranean area
Central
Anatolia
Black Sea
Eastern and
Southeastern
Anatolia
Gini index in
1968
0.559
Gini index in
1973
0.478
Gini index in
1987
0.398
Gini index in
1994
0.488
0.530
0.554
0.394
0.423
0.533
0.475
0.402
0.412
0.553
0.621
0.522
0.494
0.346
0.418
0.414
0.360
51
Table 20: Change in between regions inequality
Components of
Between Regions
Inequality Change
Between regions
Inequality in
period of origin
Between regions
inequality in final
period
Total Change
Component due to
change in
population shares
of various regions
Component due to
change in relative
income of various
regions
1968-1973 period
1973-1987 period
1987-1994 period
1968-1994 period
0.1138
0.078
0.0913
0.1137
0.0780
0.091
0.1130
0.1131
-0.0356
-0.0465
0.013
0.032
0.0218
0.0205
-0.0006
0.0059
0.0108
-0.019
0.0012
-0.0065
52
Table 21: Change in within regions inequality
Components of
Within Regions
Inequality Change
Total Change
Within regions
Inequality in
period of origin
Within regions
inequality in final
period
Component due to
change in
population shares
of various regions
Component due to
change in relative
income of various
regions
Change in within
(each) region Gini
index
1968-1973 period
1973-1987 period
1987-1994 period
1968-1994 period
-0.0013
0.0414
0.0018
0.0401
0.0016
0.0419
0.0021
0.0414
0.0401
0.0419
0.0435
0.0435
0.0011
0.0015
0.0020
0.0049
-0.0007
0.0034
0.0001
0.0025
-0.0017
-0.0032
-0.0005
-0.0053
53
V. Concluding comments:
This paper attempted to analyze the impact of internal migration on spatial inequality
in Turkey via its effect on the structure of output, the composition of income, the size
of the households and the relative importance of the various regions. In 1994 in
Turkey 72% of the household heads in urban areas were Wage and Salary Earners or
Daily Workers and 28$ were Proprietors while the corresponding proportions in rural
areas were 36% and 64%. Although Wage and Salary Earners and Daily Workers had
a lower income in rural than urban areas, the difference was much greater for
Proprietors since the latter earned 95% of the national average income in rural areas
but 225% in urban areas. For these Proprietors other sources of income (other than
income from primary or secondary job) represented 24% of their total income in
urban but only 8% in rural areas. All these data clearly imply that internal migration,
by modifying the relative importance of urban and rural areas in Turkey as a whole
and in the various regions, will have an impact on spatial inequality.
The present investigation indicated also that internal migration from rural to urban
areas and between regions induces an increase in the inequality of per capita income,
whether one looks at inequality between households or individuals, because first the
inequality of total household income is higher in urban areas, second that of the size
of households is higher in rural areas.
Finally a comparison of the 1987 and 1994 data showed that the impact of interregional migration on inequality between and within regions was maily the
consequence of changes in the (population) weight of the various regions rather than
in that of their relative income or in the inequality within each region (in the case of
overall within regions inequality).
54
References
Adelman, I. and S. Robinson, "Income Distribution and Development," in H. Chenery
and T.N. Srinivasan, eds., Handbook of Development Economics, Volume II,
Amsterdam: Elsevier Science Publishers, 1989, 951-1003.
Anand, S. and S. M. R. Kanbur, “The Kuznets Process and the InequalityDevelopment Relationship,” Journal of Development Economics 40(1993):
25-52.
Buhmann, B., L. Rainwater, G. Schmaus and T. Smeeding, “Equivalence Scales,
Well-Being, Inequality and Poverty: Sensitive Estimates Across Ten Countries
Using the Luxembourg Income Study (LIS) Database,” The Review of Income
and Wealth 34(1988): 115-42.
Coulter, F. A. E., F. A. Cowell and S. P. Jenkins, “Equivalence Scale Relativities
and the Extent of Inequality and Poverty,” Economic Journal 102 (1992):
1067-1082.
Danziger, S. and M. K. Taussig, “The Income Unit and the Anatomy of Income
Distribution,” Review of Income and Wealth 25 (1979): 365-375.
Deutsch, J. and J. Silber, "The Decomposition of Inequality by Population
Subgroups and the Analysis of Interdistributional Inequality," in J. Silber (ed.),
Handbook on Income Inequality Measurement, Kluwer Academic Publishers,
1999: 363-403.
Deutsch, J. and J. Silber, “The Kuznets Curve and the Impact of Various Income
Sources on the Link Between Inequality and Development,” with J. Deutsch,
forthcoming in the Review of Development Economics.
Fields, G. S., “A Welfare Economic Approach to Growth and Ddistribution in the
Dual
Economy,” Quarterly Journal of Economics (1979).
Fields, G.S., Poverty, Inequality and Development, New York: Cambridge University
Press, 1980.
Fields, G. S., Distribution and Development. A new look at the developing world, The
Russell Sage Foundation and the MIT Press, New York and Cambridge,
Massachusetts, 2001.
55
Gürsel, S., Levent, H., Selim R. and Sarica, Ö., 2000, Individual Income
Distribution
in Turkey, Executive Summary, Turkish Industrialists’ and Businessmen’s
Association, Istanbul.
Knight, J. B., “Explaining Income Distribution in Less Developed Countries: A
Framework and an Agenda,” Bulletin of the Oxford Institute of Economics and
Statistics (1976).
Kuznets, S., "Economic Growth and Income Inequality," American Economic Review 65
(1955):1-28.
Özmucur, S. and J. Silber “Income Inequality Decomposition by Area of
Residence (rural versus urban areas) and Income Source: The Case of Turkey
in 1987”, mimeo, 1995.
Özmucur, S. and J. Silber, 2000, “Decomposition of Income Inequality: Evidence
from Turkey,” with S. Ozmucur, Topics in Middle Eastern and North African
Economies, electronic journal, Vol. 2, Middle East Economic Association and
Loyola University Chicago, September 2000,
http: //www.luc.edu/publications/academic/
Robinson, S., "A Note on the U Hypothesis Relating Inequality and Economic
Development," American Economic Review 66 (1976): 437-440.
Selim, R. and A. McKay, 2001, “The Changes in Income Poverty in Turkey over a
period of economic liberalization,” paper presented a the Workshop on
Poverty
and Governance in the Middle East and North African Region, August 2-3,
Yemen.
Shorrocks, A. F., 1982, “Inequality Decomposition by Factor Components,”
Econometrica 50(1): 193-211.
Shorrocks, A. F., 1983, “The Impact of Income Components on the Distribution of
Family Incomes,” Quarterly Journal of Economics 98(2): 311-26.
Silber, J. “Factor Components, Population Subgroups and the Computation of the
Gini Index of Inequality”, Review of Economics and Statistics, 71, 1989, pp
107-115.
56
Appendix A: Some Additional Tables
Table A-1: The regional income distribution in Turkey in 1968
Region
Central Anatolia
Black Sea
Marmara and
Aegean Area
Mediterranean
Area
Eastern Anatolia
Ankara
Istanbul
Izmir
Share of region in Gini index of
household
total Turkish
household income incomes
Share of region in
total Turkish
number of
households
0.182
0.177
0.235
0.160
0.147
0.187
0.549
0.553
0.449
0.153
0.114
0.530
0.138
0.044
0.052
0.020
0.115
0.071
0.136
0.070
0.621
0.368
0.488
0.622
Summary indicators
Gini within areas = 0.075
Gini between areas = 0.185
Overlap between regional income distributions = 0.305
Gini for whole Turkey = 0.565
57
Table A-2: The regional income distribution in 1973
Region
Central Anatolia
Black Sea
Marmara and
Aegean Area
Mediterranean
Area
Eastern Anatolia
Ankara
Istanbul
Izmir
Share of region in
total Turkish
number of
households
0.175
0.149
0.233
Share of region in Gini index of
household
total Turkish
household income incomes
0.180
0.158
0.208
0.486
0.522
0.454
0.152
0.132
0.554
0.147
0.044
0.085
0.019
0.099
0.054
0.139
0.030
0.494
0.424
0.454
0.506
Summary indicators
Gini within areas = 0.075
Gini between areas = 0.128
Overlap between regional income distributions = 0.302
Gini for whole Turkey = 0.505
58
Table A-3: The regional income distribution in 1987
Region
Marmara and
Aegean area
Mediterranean area
Central Anatolia
Black Sea
Eastern and
Southeastern
Anatolia
Share of region in
total Turkish
number of
households
0.370
Share of region in Gini index of
household
total Turkish
household income incomes
0.450
0.398
0.134
0.107
0.394
0.243
0.106
0.147
0.215
0.089
0.139
0.402
0.346
0.418
Summary indicators
Gini within areas = 0.105
Gini between areas = 0.091
Overlap between regional income distributions = 0.220
Gini for whole Turkey = 0.416
59
Table A-4: The regional income distribution in 1994
Region
Marmara
Agean area
Mediterranean
area
Central Anatolia
Black Sea
Eastern Anatolia
Southeastern
Anatolia
Share of region in
total Turkish
number of
households
0.266
0.157
0.125
Share of region in Gini index of
household
total Turkish
household income incomes
0.386
0.139
0.111
0.490
0.401
0.423
0.179
0.128
0.071
0.074
0.154
0.109
0.057
0.045
0.412
0.414
0.342
0.351
Summary indicators
Gini within areas = 0.085
Gini between areas = 0.139
Overlap between regional income distributions = 0.240
Gini for whole Turkey = 0.464
60
Appendix B: The Macroeconomic Environment in Turkey
between 1960 and 20005.
The primary goals of economic policies of the 1960s and 1970s was the protection of
the domestic market and rapid industrialization (Owen & Pamuk, 2000) Five-year
plans and annual programs, which were binding for the public sector but indicative for
the private sector, were used to coordinate investment decisions for long-term growth.
The state played a major role by putting restrictions on imports (direct quotas or high
tariffs), undertaking major investment projects, determining prices of major factors
and goods and services, and giving subsidies. This role was necessary to support an
import substitution industrialization (ISI) policy. In this framework, monetary policy,
which was generally geared towards short-term goals, played a secondary role. The
Central Bank governor was regarded as just another civil servant accountable to the
Ministry of Finance.
During most of this period, the private sector operating in a highly protected
environment made handsome profits. This protection enabled the private sector to
achieve a major shift from producing food and clothing to producing cars, household
durables and electronic goods. However, because of the large domestic market and
relatively low quality of goods, exports of manufactured goods were almost
nonexistent. The required foreign exchange to import necessary raw materials and
machinery came from workers’ remittances and agricultural exports (hazelnuts, figs,
cotton). During this period the growth rate of the GDP was close to 7 percent, while
5
See Alper & Onis (2001), Aricanli & Rodrik(1990), Boratav & Yeldan (2001), Boratav, Yeldan,
&Kose (2000), Cecen, Dogruel & Dogruel (1994) Cizre-Sakallioglu & Yeldan (2000, Dibooglu &
Kibritcioglu (2001), Ertugrul & Selcuk (2001), Metin_Ozcan, Voyvoda & Yeldan (1999), Onis (2000),
Onis & Riedel (1993), Owen&Pamuk(2000), Rittenberg (1998), Rodrik (1991), and Yeldan (2000).
61
the growth rate in the manufacturing sector exceeded 10 percent. As a result of these
favorable
developments on the economic front and new rights under the 1961
Constitution real wages were doubled. In rural areas, farmers benefited also from this
expansion of the domestic market as well as from agricultural price support programs,
subsidies for fertilizers and low-interest credits for tractors. Governments were also
able to control the domestic terms of trade in favor of agriculture, especially before
elections.
The growth in Europe also helped Turkey. The number of “guest workers” in Europe
increased, and by 1973 workers’ remittances exceeded export revenues and reached
almost 5% of the GDP. This major inflow, however, contributed to an overvaluation
of the Lira, and increased the demand for imports while reducing the demand for
exports. Intermediate goods which were to be imported for industrial production
became rather expensive, especially after the first oil price shock in late 1973. Instead
of taking drastic measures (similar to the ones taken in Europe) to save energy,
populist coalition governments continued to adopt expansionary policies. The
consequences of these policies were a depletion of foreign exchange reserves and a
significant increase in foreign debt. A new instrument created by the government,
“convertible Lira deposit”, allowed private firms to borrow abroad, with an exchange
rate guarantee, at the expense of the state. In 1978 Turkey had its most severe balance
of payments crisis. The IMF demanded the elimination of import controls, a
devaluation of the Lira, and cutbacks in government subsidies. The coalition
government was not ready to bear the political consequences of such measures.
Foreign exchange and price controls adopted by the government, and the second oil
62
price shock resulted in shortages. The monetary expansion to finance budget deficits
resulted in a rising inflation so that the rate of inflation was three digits in 1980.
The new coalition government had to announce a radical policy package in January of
1980. Its goals were to lower the rate of inflation and to create an export-oriented and
liberalized economy. This required a major devaluation, the elimination of price
controls and subsidies, a liberalization of the trade regime and of the banking sector,
and reductions in real wages and agricultural incomes. The coalition government was
willing to adopt these policies, but did not have the political support to implement
them. Political and social unrest was the reason for the military coup of September
1980. The military government, which promised to have general elections in 1983 and
kept that promise, adopted this program, prohibited labor union activity and put
controls over wage increases. The program was successful in lowering the rate of
inflation from over 100 percent in 1980 to below 30 percent in 1983, and doubling
exports during the same period. It was also successful in lowering the shares of
agricultural incomes and wages in national income. Since the program had the full
support of the IMF and the World Bank, external debt was rescheduled and foreign
exchange problems were practically ended. The full support of the program by
international organizations, despite a military government, was largely due to the
revolution in Iran and the strategic location of Turkey.
This program however was not as successful in other areas. Because of higher interest
rates and lower credits, private investment was affected adversely. As a result the
growth of the GDP was not as impressive as during the ISI period. Foreign debt which
accumulated at a much higher pace, acted as a drag on the public sector. The principal
63
and interest payments on debt represented about fifty percent of the entire budget,
which made it difficult for the government to adopt any social program. Wages and
agricultural prices were kept down until the general elections of 1987. In urban areas,
wage and salary earners started to moonlight in the “informal sector”. In rural areas,
agricultural output failed to keep pace with population growth. Migration from rural
to urban metropolitan areas gained momentum6. On the other hand, a new class of
“rentiers” emerged living from high interest rates on bank deposits (domestic and
foreign currency). Before the general elections of 1987 public sector wages and
salaries and agricultural incomes were increased sharply. Similar policies were
adopted in 1989 at the time of the election of the President by the Parliament. By the
end of the decade, the economy was vulnerable to another crisis.
In 1989 the capital account was fully liberalized. The need to finance the increasing
public sector deficits was also an important motivation to liberalize the capital
account. Under the new regime, short-term capital inflows became the ultimate
financing source of the fiscal deficit. As the government tried to fix the real exchange
rate and the real interest rate to maintain the international competitiveness and to
attract foreign capital, with persistent public sector deficits, the result was high and
variable inflation, which led to further volatility in real interest rates and real
exchange rates.
The Gulf War made things much worse for Turkey in the 1990s. Oil pipeline
revenues, close to a billion US Dollars ended, as well as cross-border trade with
neighbors. The livelihood of the people living close to the border were adversely
6
See Celasun (1986, 1989), Mutlu (1989, 1998) and Tekeli (1981). See also Shoter (1995) on
population.
64
affected. This was another reason for a rapid migration to metropolitan areas in the
region and to the West of the country. This rapid movement to urban areas increased
the cost to municipalities. The debt of municipalities also increased during this period.
An attempt to fix the interest rate, despite large fiscal deficit and high inflation, led to
another crisis in 1994. The growth rate of the GDP was negative for the first time
after the crisis of 1980. With the help of the new stabilization program, the economy
was quick to recover, but the damage was already done. Turkey had economic
difficulties a few years later. IMF supported stabilization programs were implemented
in 1998 and 1999. Despite IMF supervision, the economy had a banking crisis in
November 2000 and a currency crisis in February 2001. The Lira had to float after
February.
A crisis with a negative growth rate and a high rate of inflation always leads naturally
to a more unequal distribution of income. Since, the government is still the biggest
player in the economy (biggest employer, producer, and policy maker), the economic
policies adopted had a direct effect on the level of inequality. There was for example a
deliberate policy of keeping real wages under control but the policy aiming at
increasing public revenues by increasing the price of oil, irrespective of international
prices, had evidently also an impact on inequality. Similarly the value-added tax
which was implemented in 1985 with the dual goal to increase government revenue
and to reduce tax evasion ended up being a regressive tax which contributed to
inequality in after-tax incomes. It is not easy to distinguish the contribution of each of
these policies to income inequality, but it should be clear that, given the dominance of
the state in the economic decision-making process and its active participation in the
65
production and service sectors, government policies must have an impact on income
distribution so that inequality is likely to have increased also in 2001.
66
TABLE B1: MACROECONOMIC INDICATORS
Exports/GD (Exports+imports)/G
GDP Inflatio Unemployme Share Current
P ratio (%) DP ratio (%)
account
nt Rate
of
growt n
wages balance/GD
h rate
P ratio (%)
in
nationa
l
income
195 9.4
-5.5
1.4
21.3
-1.4
7.6
15.9
0
195 12.8 -1.0
1.7
19.8
-2.3
7.6
17.2
1
195 11.9 5.8
1.8
20.6
-4.1
7.6
19.2
2
195 11.2 3.7
2.8
20.9
-2.9
7.1
16.7
3
195 -3.0 10.0
3.1
23.3
-3.1
5.9
14.3
4
195 7.9
8.2
3.0
22.4
-2.6
4.6
11.9
5
195 3.2
14.3
3.1
22.8
-1.0
3.9
9.0
6
195 7.8
11.8
2.7
20.7
-0.6
3.3
7.1
7
12.5
2.8
21.0
-0.5
2.0
4.5
195 4.5
8
195 4.1
26.0
2.8
22.2
-0.9
2.3
5.3
9
196 3.4
5.6
3.1
22.1
-2.7
6.2
15.2
0
196 2.0
3.8
3.4
24.4
-3.1
6.3
15.6
1
196 6.2
3.6
3.3
22.7
-3.8
6.0
15.7
2
196 9.7
9.8
3.3
22.8
-4.0
5.0
14.2
3
196 4.1
0.2
3.5
24.7
-1.4
5.2
12.0
4
196 3.1
4.6
3.6
27.6
-0.9
5.4
12.2
5
196 12.0 8.4
3.6
27.7
-1.6
4.8
11.9
6
196 4.2
14.1
4.7
28.9
-1.0
4.6
10.7
7
196 6.7
6.2
5.1
29.5
-1.2
2.7
6.9
8
196 4.3
7.8
5.8
29.1
-1.1
2.6
6.6
9
197 4.4
8.1
6.3
29.9
-0.9
3.2
8.4
0
197 7.0
16.5
6.6
30.2
-0.6
3.9
10.6
1
197 9.2
13.7
6.2
29.4
0.0
3.9
10.9
2
197 4.9
16.0
6.6
33.4
1.7
4.6
11.9
3
197 3.3
18.6
7.1
29.5
-1.8
3.9
13.6
4
197 6.1
19.8
7.4
28.8
-3.4
2.9
12.3
5
67
Percentag
e
Changes
in
Lira/USD
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
221.4
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
0.0
26.0
32.1
-6.5
0.0
-1.9
4.5
197
6
197
7
197
8
197
9
198
0
198
1
198
2
198
3
198
4
198
5
198
6
198
7
198
8
198
9
199
0
199
1
199
2
199
3
199
4
199
5
199
6
199
7
199
8
199
9
200
0
200
1
9.0
16.4
8.7
30.9
-3.7
3.6
12.5
10.8
3.0
28.0
9.8
31.9
-5.0
2.8
11.6
11.6
1.2
47.2
9.8
33.1
-1.9
3.4
9.8
36.6
-0.5
56.8
8.6
31.0
-1.6
2.5
7.9
31.9
-2.8
115.6
8.1
27.6
-4.7
4.0
14.4
128.0
4.8
33.9
7.1
27.4
-2.8
6.9
19.5
61.2
3.1
21.9
7.0
26.0
-1.4
9.0
21.9
37.1
4.2
31.4
7.7
24.8
-3.1
9.5
23.8
38.7
7.1
48.4
7.6
22.9
-2.4
12.2
29.2
63.2
4.3
45.0
7.1
23.1
-1.5
12.1
28.6
42.2
6.8
34.6
7.9
25.1
-1.9
9.9
23.9
28.9
9.8
38.9
8.3
24.4
-0.9
11.8
27.3
28.1
1.5
77.6
8.4
25.4
1.8
13.2
28.4
67.0
1.6
63.2
8.6
28.0
0.9
10.8
25.6
48.1
9.4
60.3
8.0
31.8
-1.7
8.5
23.4
22.8
0.3
66.1
7.9
37.4
0.2
9.0
22.8
60.2
6.4
70.1
8.0
37.2
-0.6
9.2
23.5
64.6
8.1
66.4
7.7
36.3
-3.6
8.5
24.9
60.5
-6.1
106.3
8.1
30.2
2.0
13.9
31.0
169.9
8.0
88.6
6.9
26.5
-1.4
12.6
32.9
53.5
7.1
80.3
6.0
28.5
-1.3
17.4
40.6
77.9
8.3
86.0
6.7
31.1
-1.4
16.6
41.3
86.9
3.9
84.7
6.8
30.7
1.0
15.0
36.9
71.7
-6.1
65.0
7.6
37.4
-0.8
15.5
36.7
60.9
6.3
54.6
6.6
35.8
-4.9
15.2
41.7
48.5
-9.4
54.4
8.5
35.7
2.3
23.5
50.1
96.5
68
Sources:
GDP growth rate: State Institute of Statistics,
Inflation rate : 1950-1968 - Treasury, 1969-2001- State Institute of Statistics,
Unemployment rate: 1950-1987 – Bulutay (1995), 1988-2001 – State Institute of
Statistics,
Share of wages in national income: 1950-1986 – Ozmucur (1996), 1987-2001 State Institute of Statistics,
Current account balance – State Institute of Statistics and Central Bank, Exports
– State Institute of Statistics and Central Bank
Exports +imports – State Institute of Statistics and Central Bank
Turkish Lira/USD – Central Bank
69
Table . Urban and Rural Population in Turkey
year
Total Popunnual growt Urban Popnnual growt Rural Popunnual growt share of urb
(million)
(million)
(million)
1927
13.648
3.306
10.342
24.2
1935
16.158
2.11
3.803
1.77 12.355
2.25
23.5
1940
17.821
1.96
4.346
2.71 13.475
1.75
24.4
1945
18.790
1.06
4.687
1.52 14.103
0.92
24.9
1950
20.947
2.17
5.244
2.27 15.703
2.17
25.0
1955
24.065
2.78
6.927
5.72 17.138
1.76
28.8
1960
27.755
2.85
8.860
5.05 18.895
1.97
31.9
1965
31.391
2.46
10.806
4.05 20.585
1.73
34.4
1970
35.605
2.52
13.691
4.85 21.914
1.26
38.5
1975
40.348
2.50
16.869
4.26 23.479
1.39
41.8
1980
44.737
2.07
19.645
3.09 25.092
1.34
43.9
1985
50.664
2.49
26.866
6.46 23.798
-1.05
53.0
1990
56.473
2.17
33.656
4.61 22.817
-0.84
59.6
2000
67.804
1.83
44.006
2.72 23.798
0.42
64.9
Note: population censuses were conducted in October
TABLE B2. Average Annual Growth Rates in Population - Geographic Regions (%)
1945-1950
1950-1955
1955-1960
1960-1965
1965-1970
1970-1975
1975-1980
1980-1985
1985-1990
1990-2000
Turkey
Marmara Agean
Black
Sea
Mediterranean Central
Anatolia
South
Eastern
Anatolia
Eastern
Anatolia
2.17
2.77
2.85
2.46
2.52
2.50
2.07
2.49
2.17
1.83
1.47
3.18
2.95
2.38
3.17
3.30
3.14
3.25
3.61
2.67
1.90
1.84
2.57
2.12
1.59
1.35
1.20
1.22
0.36
0.37
2.67
3.42
3.38
2.80
2.96
3.45
2.87
3.05
2.75
2.14
2.78
5.52
2.96
2.81
3.38
2.73
2.10
3.75
3.62
2.48
2.71
1.47
2.68
2.63
2.45
2.15
1.47
1.77
0.51
1.38
2.13
2.65
2.72
2.27
2.03
2.07
1.97
2.50
2.37
1.63
2.42
3.34
2.55
2.62
2.60
2.63
1.58
2.19
1.52
1.58
2000
67.803
17.365
8.939
8.439
8.706
11.609
6.609
6.137
population
(million)
2000 share 100.0
25.61
13.18
12.45
12.84
17.12
9.75
9.05
(%)
___________________________________________________________________________________
__
Source : State Institute of Statistics
Table B3. Income Distribution in Turkey, 1963- 1994
70
Turkey
lowest 20%
second 20%
middle 20%
fourth 20%
top 20%
1963
4.5
8.5
11.5
18.5
57.0
1968
3.0
7.0
10.0
20.0
60.0
1973
3.5
8.0
12.5
19.5
56.5
1987
5.2
9.6
14.1
21.2
49.9
1994
4.9
8.6
12.6
19.0
54.9
Gini
0.550
0.560
0.515
0.437
0.492
lowest 40%
13.0
10.0
11.5
14.9
13.5
top /bottom 20% 12.7
20.0
16.1
9.5
11.3
Kuznets
0.463
0.500
0.456
0.389
0.436
_____________________________________________________________________
___
Sources: 1963- Cavusoglu & Hamurdan (1966), 1968 – Bulutay, Serim, Timur
(1970), 1973 – Devlet Planlama Teskilati (1976), 1987 – Devlet Istatistik Enstitusu
(1990), 1994 – Devlet Istatistik Enstitusu (1997)
71
Table B4. Income Distribution in Urban and Rural Turkey, 1973-1994
Urban
lowest 20%
second 20%
middle 20%
fourth 20%
top 20%
1973
1987
5.4
9.3
13.6
20.7
50.9
1994
4.8
8.2
11.9
17.9
57.2
Gini
0.479
0.444
0.515
Rural
lowest 20%
second 20%
middle 20%
fourth 20%
top 20%
1973
1987
5.2
10.0
15.0
22.0
47.8
1994
5.6
10.1
14.8
21.8
47.7
Gini
0.543
0.417
0.414
_____________________________________________________________________
Sources: 1973 – Devlet Planlama Teskilati (1976), 1987 – Devlet Istatistik Enstitusu
(1990), 1994 – Devlet Istatistik Enstitusu (1997)
72
Appendix C: More on the Data Sources on the Distribution of Incomes
in Turkey (1963-1994)7
The 1963 income distribution study was the first income distribution study carried out
by the State Planning Organization (Cavusoglu & Hamurdan, 1966). Incomes were
gathered from the income tax revenues of 327000 taxpayers. The study was criticized
on the grounds that income taxes had a very limited coverage.
According to
Cavusoglu & Hamurdan (1966), the lowest twenty percent of households received 4.5
percent, the second lowest 20% receives 8.5 percent. The share of the middle group
was 11.5 percent. The top two groups received 18.5 and 57 percent of incomes (Table
1). Boratav (1966) claimes that inequality was underestimated because of tax evasion,
while Sarc (1967, 1970) argues the opposite.
The 1968 study was based on a survey carried out by Institute of Demographic
Studies of Hacettepe University (Bulutay, Ersel, Timur, 1971). Disposable income
data used in the calculations were based on the replies given by married males whose
wives were less than 45 years of age in each household. The households with no
married women younger than 45 years of age were excluded from the study. Those
households constitute 17.2 percent of the total in Turkey. The reason for this
exclusion was due to the initial goal of the survey, which was to look at demographic
data (fertility in particular) rather than at income distribution. After having talked to
one of those who launched this survey (Timur) the authors, Krzyzaniak and Ozmucur
7
See Boratav (1966, 1969, 1990), Bulutay & Ersel(1967), Bulutay, Timur, & Ersel (1971)
,Celasun(1986, 1989), Cavusoglu & Hamurdan (1966) , Dervis & Robinson (1980), Esmer, Fisek
&Kalaycioglu (1986), Hansen (1991), Herslag (1990), Gursel, Levent, Selim, & Sarica (2000),
Kasnakoglu (1978,1997),Kazgan, Onder, Kirmanoglu & Tuncer (1992),Krzyzaniak & Ozmucur (1973)
,Özbudun & Ulusan (1980) ,Özmucur (1996),Sarc (1966, 1967, 1970),Sonmez (1971), Sonmez (1990),
State Planning Organization (1976), State Institute of Statistics (1979, 1982,1990,1997), Tansel (1992),
Varlier (1982)., and World Bank (2000).
73
(1973) made a correction for excluded households and got an estimate of the Gini
coefficient equal to 0.57, which is slightly higher than the one in the original study.
The 1973 study was also based on a country-wide survey carried out by the
Demographic Studies Institute of Hacettepe University (State Planning Organization,
1976). In addition to five geographical regions, the cities of Istanbul, Ankara and
Izmir were considered as metropolitan areas. The regions remaining outside the
metropolitan areas were divided into urban and rural areas. The Gini coefficient was
found to be equal to 0.51. Dervis and Robinson (1980) argued that non-agricultural
incomes and agricultural population in the survey were underestimated, which led to a
downward bias in equality. Their estimate of the Gini coefficient was slightly lower
(0.50). Agricultural incomes had a Gini coefficient of 0.56 compared to 0.45 for the
non-agricultural incomes. In the original study the corresponding coefficients were
0.60 and 0.63, respectively. According to Celasun (1986) the Gini estimates were 0.51
for the total, 0.57 for the agricultural, and 0.43 for the non-agricultural incomes.
The 1987 study was the first survey covering Turkey as a whole (State Institute of
Statistics, 1990). Income definitions included personal disposable income (actual
payments made e.g. salaries, interest, profit, rent and unilateral transfers from the
public and private enterprises and from abroad) and income in kind. The Gini
coefficient was equal to 0.43. Such an estimate, the lowest coefficient obtained in
income distribution surveys, was questioned by many researchers8. Celasun (1989)
reported also preliminary results of this survey. They indicated that the top 20 percent
had received 55 percent of the total income, and bottom 40 percent have received 11
74
percent of total income. These figures are very different from the ones reported in the
final report. Esmer, Fisek, Kalaycioglu (1986) obtained a Gini coefficient of 0.50,
which is in line with the 1973 results.
The 1994 study covered also Turkey as a whole and was conducted by the same
organization (State Institute of Statistics, 1997). These two survey results are
therefore probably the most comparable. Income definitions included personal
disposable income (actual payments made e.g. salaries, interest, profit, rent and
unilateral transfers from the public and private enterprises and from abroad) and
income in kind. The Gini coefficient was found to be equal to 0.49 and this was
significantly higher than the one obtained in the 1987 survey.
8
The State Institute of Statistics has conducted consumer expenditure surveys in rural areas in 1973-74
(SIS, 1979) and urban areas in 1978-79 (SIS, 1982). Gini coefficients obtained from consumer
expenditure surveys were 0.47 and 0.40, respectively.
75
Additional Bibliography for Appendices B and C.
Alper, E. and Z. Onis, “Financial Globalization, the Democratic Deficit and Recurrent
Crises in Emerging Markets: The Turkish Experience in the Aftermath of Capital
Account Liberalization”, July 2001, mimeo.
Aricanli, T. and D. Rodrik, eds. The Political Economy of Turkey, Debt, Adjustment and
Sustainability, St. Martin’s Press, New York, 1990.
Boratav, K., “Turkiye’de Kisisel gelir Dagilimi ve Devlet Planlama Teskilatinin
Arastirmasi”, Ankara Universitesi Siyasal Bilgiler Fakultesi Dergisi, XXI (1966), 45102.
Boratav, K., 100 Soruda gelir Dagilimi (Kapitalist Sistemde, Turkiye’de ve Sosyalist
Sistemde), Gercek Yayinevi, Istanbul, 1969.
Boratav, K.,”Inter-class and intra-class Relations of Distribution under ‘Structural
Adjustment’: Turkey During the 1980’s” in Aricanli, T. and D. Rodrik, eds. The
Political Economy of Turkey, Debt, Adjustment and Sustainability, St. Martin’s Press.
New York, 1990.
Boratav, K. and E. Yeldan, Turkey, 1980-2000: Financial Liberalization,
Macroeconomic (In)-Stability, and Patterns of Distribution, December 2001, mimeo.
Boratav, K., E. Yeldan, and A. Kose, Globalization, Distribution and Social Policy:
Turkey, 1980-1998, CEPA (Center for Economic Policy Analysis, New School,
February 2000, mimeo.
Bulutay, T., Employment, Unemployment and Wages in Turkey, International Labor
Office and State Institute of Statistics, Ankara, SIS Printing Division, 1995.
Bulutay, T., H. Ersel. “The Distribution of Income in Certain Cities”, Milletlerarasi
Munasebetler Turk Yilligi, 1967.
Bulutay, T., S. Timur, and H.Ersel, Turkiye’de Gelir Dagilimi, 1968, Ankara
Universitesi Siyasal Bilgiler Fakultesi Yayinlari No: 325, Ankara.
Cecen, A., S. Dogruel and F. Dogruel, “Economic Growth and Structural Change in
Turkey, 1960-1988, International Journal of Middle East Studies, Vol. 26, Issue 1,
February 1994, pp.37-56.
Celasun, M., “Income Distribution and Domestic terms of Trade in Turkey, 19781983”, METU Studies in Development, 13 (1986), 193-216.
Celasun, M., “Income Distribution and Employment Aspects of Turkey’s Post 1980
Adjustment”, METU Studies in Development, 16 (1989), 1-31.
76
Cizre-Sakallioglu, U and E. Yeldan, “Politics, Society and Financial
Liberalization:Turkey in the 1990s”, Development and Change, Vol. 31, 2000, pp.
481-508.
Cavusoglu, T. and Y. Hamurdan, Gelir Dagilimi Arastirmasi, 1963, Devlet Planlama
Teskilati Yayin No. 500. Ankara.
Dervis, K. and S. Robinson, “The Structure of Income Inequality in Turkey (195073)” in Ozbudun and Ulusan, eds. The Political Economy of Income Distribution in
Turkey, Holmes and Meier Publishers, New York, 1980.
Deutsch, J. and J. Silber, “The Kuznets Curve and the Impact of Various Income
Sources on the Link Between Inequality and Development,” mimeo, 1999.
Dibooglu, S. and A. Kibritcioglu, Inflation, Output, and Stabilization in a High
Inflation Economy: Turkey, 1980-2000. University of Illinois at Urbana-Champaign,
Office of Research Working Paper No: 01-0112. April, 2001, mimeo.
Ertugrul, A. and F. Selcuk, “A Brief Account of the Turkish Economy, 1980-2000”,
Russian and East European Finance and Trade, 2001.
Esmer, Y. , H. Fisek, E.Kalaycioglu, Turkiye’de Sosyo Ekonomik Oncelikler, Hane
Gelirleri, Harcamalari ve Sosyo-Ekonomik Ihtiyaclar Uzerine Arastirma Dizisi, Cilt
II. TUSIAD, Istanbul, 1986.
Frieden, J. Inequality, Causes and Possible Futures, April 2001, mimeo
Hansen, B. The Political Economy of Poverty, Equity and Growth: Egypt and Turkey,
World Bank Comparative Studies, Oxford University Press, 1991
Herslag, Z.Y.. “Growth, development, Equity-A Case Study of Turkey”, Bogazici
University Journal of Economics and Administrative Studies (Demirgil Memorium),
IV, 25-34.
Gursel, S., H. Levent, R. Selim, and O. Sarica, Individual Income Distribution in
Turkey, A Comparison With the European Union, TUSIAD Publication NO: T/200012/296, December 2000.
Kasnakoglu, Z. “A Simultaneous Model Approach to the Determinants of Male
Earnings Differentials in Turkey for 1968”, The Review of Economics and Statistics,
LX (1978), 307-312.
Kasnakoglu, Z. “Income Distribution in Turkey, Who Gets What?”, Private Viev,
Autumn 1997.
Kazgan, G., I. Onder, H. Kirmanoglu, N. Tuncer, Turkiye’de Gelir Dagilimini Bozan
Etkenler ve Iyilestirilmesine Iliskin Politikalar, TOBB, Ankara, 1992.
Krzyzaniak, M. and S. Ozmucur, “The Distribution of Income and the Short-run
Burden of Taxes in Turkey, 1968”, Finanzarchiv, 32 (1973), 69-97.
77
Metin_Ozcan, K., E. Voyvoda, and E. Yeldan, Dynamics of Macroeconomi
Adjustment in A Globalized Developing Economy: Growth, Accumulation and
Distribution, Turkey, 1969-1998. July 1999, mimeo.
Mutlu, S., “Population and Agglomeration Trends in the Turkish Settlement System:
An Empirical Analysis and Some of its Implications”, METU Studies in
Development, Vol. 16 (1-2), 1989, pp. 99-125.
Mutlu, S., “Turkiye’de Mer’alar, Hayvanlar ve Insanlar”, METU Studies in
Development, Vol. 25 (3), 1998, pp. 447-490.
Owen, R. and S. Pamuk, A History of the Middle East Economies in the Twentieth
Century, Harvard University Press, 2000.
Önis, Z. Turkish Economy at the Turn of a New Century: Critical and Comparative
Perspectives, February 2000, mimeo.
Önis, Z. and J. Riedel, Economic Crises and Long-term Growth in Turkey, World
Bank Publications, Washington, D.C., 1993.
Özbudun, E. and A. Ulusan, eds. The Political Economy of Income Distribution in
Turkey, Holmes and Meier Publishers, New York and London, 1980.
Özmucur, S. Turkiye’de Gelir Dagilimi, Vergi Yuku ve Makroekonomik Gostergeler,
Bogazici Universitesi Yayinlari, Istanbul, 1996.
Rittenberg, L. ed., The Political Economy of Turkey in the Post-Soviet Era:Going West
and Looking East, Praeger Publishers, Connecticut, 1998.
Rodrik, D., “Premature Liberalization, Incomplete Stabilization:The Ozal Decade in
Turkey”, in Bruno, et.al. Lessons of Economic Stabilization and its Aftermath, MIT
Press, Cambridge, 1991.
Sarc, O.C., “Kalkinma Plani ve Milli Gelirimizin Dagilimidaki Degismeler”, Iktisat
Fakultesi Mecmuasi, 35 (1966), No.1-2.
Sarc, O.C., Devlet Planlama teskilati Arastirmasi, Iktisadi Arastirmalar Tesisi No. 16,
Dogan Kardes Matbaasi, Istanbul, 1967.
Sarc, O.C., Gelir Dagilimi (Disarda ve Turkiye’de), Ekonomik ve Sosyal Etudler
Konferans Heyeti, Celtut Matbaacilik, Istanbul, 1970.
Shorter, F. C., “The Crisis of Population Knowledge in Turkey”, New Perspectives on
Turkey, Vol. 12, Spring 1995, pp. 1-31.
Sonmez, A., “Literature on Income Distribution in Turkey”, METU Studies in
Development, 1971, 543-550.
Sonmez, M., Turkiye’de Gelir Esitsizligi, Iletisim Yayinlari, Istanbul, 1990.
78
State Planning Organization, Gelir Dagilimi, 1973, Devlet Planlama Teskilati Yayin
No. 1495, Ankara, 1976.
State Institute of Statistics, Kirsal Kesim Gelir Dagilimi Ve Tuketim harcamalari,
1973-1974, Devlet Istatistik Enstitusu Yayin No. 881, Ankara, 1979.
State Institute of Statistics, Kentsel Yerler hanehalki Gelir Ve Tuketim harcamalari
Anketi Sonuclari, 1978-1979, Devlet Istatistik Enstitusu Yayin No. 999, Ankara,
1982.
State Institute of Statistics, 1987 Hanehalki Gelir Ve Tuketim harcamalari Anketi
Sonuclari- Gelir Dagilimi, Devlet Istatistik Enstitusu Yayin No. 1441, Ankara, 1990.
State Institute of Statistics, Income Distribution Survey Results, 1994, State Institute of
Statistics Publications No: 2051, Ankara, 1997.
Tansel, A. “Household Saving, Income, and Demographic Interactions”, METU Studies
in Development, 19 (1992), 91-114.
Tekeli, I. “Dort Plan Doneminde Bolgesel Politikalar ve Ekonomik Buyumenin
Mekansal Farklilasmasi”, METU Studies in Development, Special Issue, 1981, pp. 369390.
Varlier, O. Turkiye’de Kazanc Esitsizliklerinin Nedenleri, Gazi Universitesi Yayinlari,
No. 13. TUBITAK Matbaasi, Ankara, 1982.
Yeldan, E. The Impact of Financail Liberalization and the Rise of Financail Rents
on Income Inequality: The Case of Turkey, WIDER, The United Nations
University, Working Papers No: 206, November 2000.
World Bank, Turkey, Economic Reforms, Living Standards and Social Welfare
Study, Washington, D.C., January 2000.
79