The Decomposition of Regional Income Inequalities of Turkey

The Decomposition of Regional Income Inequalities of Turkey Ayse Aylin BAYAR a a Faculty of Management, Istanbul Technical University (ITU) Abstract Even there is an economic growth since 2000s in the Turkish economy; some increasing concerns about the distributional consequences of policies implemented throughout this period are also being discussed. The inequality level of Turkey is improved in the same period, however she still possesses one of the worst income distribution compared to the other developing countries. Regions of Turkey have some disparities in terms of economic, demographic and infrastructural characteristics. In that sense, these disparities may deteriorate the overall inequality of Turkey. Data for the empirical analysis is taken from SILC (Survey of Income and Living Conditions) which is conducted by TurkStat for the selected years 2006, 2010 and 2013. At first, basic regional income inequality measures are revealed in order to explore the differences between the regions. After revealing regional income inequality levels, the overall inequality decomposed into between and within inequality in terms of regions. Empirical findings show that the overall inequality is result from the within inequality of the regions for all investigated years. These results drive the attention through to the regional income inequality of Turkey, and therefore, the one another ultimate aim is to decompose regional inequalities according to household characteristics. The subgroup decomposition of regional income inequality estimations are done for the educational attainment and the gender of the household head. It is explored that, the within inequalities are responsible for the within regional income inequality of the all regions. Keywords : Regional Income Inequality, Decomposition, Turkey JEL Classification : D31, 015, 050, R10 Corresponding author: Ayse Aylin Bayar, Faculty of Management, Department of Management Engineering, 34367 Macka Istanbul / Turkey; Fax: +90-212-240 72 60; email: bayaray@itu.edu.tr. 1

1. INTRODUCTION Inequality is one of the most substantial social problem of the economies. Especially, developing economies and the less developed ones suffer from the unequal distribution of income and the inequality in income distribution creates undesired diverse problems in the economy. During the development process, in the economies some of the income groups benefit more than the others and the shift in the distribution of the income are mostly for the good of these groups. The problem of the developing countries not rely on the fact that a specific group is getting richer, the sole problem is to redistribute the income equally within the society. Besides, within a country, the distributional effects of economic growth are not the same across the regions. Therefore, the overall inequality of the economy stays unresolved until the gap between the regions is narrowed. Therefore, there is a growing literature on the inequality and on its distributional effects to overall economy. Researchers try to solve the inequality problem and recommend many different macroeconomic policies and governments applied policies for the specific groups who hurt from the inequality more than others. In that sense, at the region base, these policies mainly target the low-income regions in order to sustain a better condition for them. Therefore, for Turkey, the concern about the inequality is one of the priorities for the government, as well. Turkey shows a great performance with regard to macroeconomic indicators. She captures an increasing trend for economic growth and there exists a good macroeconomic environment in the economy. Despite all the improvements in the economy, inequality problem still continues its existence. For sure, unequal distribution in Turkey has been improved at the same period. However, this improvement is not adequate to reduce the gap between the different income groups and/or regions. As in Turkey, the difference between the different parts of the country is also one of the reasons for the unequal distribution, in order to solve the regional income inequality problems, it is crucial to identify the contribution of regions to the overall economy. Therefore, in this paper the one aim is to investigate the inequality for different regions and test whether the overall inequality is result from the within inequality in the regions or between inequality across the regions. Besides, the determinants of the regional income inequality are also revealed in order to capture the different population subgroups effect on the regional income inequality. 2

In the present paper, Income and Living Conditions Survey (SILC) data for the years 2006 and 2013 are employed. First, the regional and overall inequality measures will be given. Afterwards, the overall inequality is decomposed into within and between subgroups in terms of regions. In order to do that, the subgroup decomposition method is applied. And at last, one more decomposition analysis is done to reveal the determinants of the regional income inequality. With doing so, the inequality of the regions will be explored. The results point out that the overall inequality in Turkey is improved from the year 2006 to 2013. Besides, the existence of the regional disparities is also captured. It is seen that some of the regions in Turkey (North-East Anatolia and Mediterranean regions) are more unequal than the others. Also, decomposition analysis of inequality shows that the overall inequality in Turkey mainly results from within regional inequality. At last, the decomposition of the regional income inequality figure out that the within inequalities for the educational attainment and the gender is responsible for the inequality in the regions. The paper is organized as follows. In Section 2 a very brief literature review about regional income inequality is given while Section 3 is dedicated the data and descriptive statistics. The methodology is described in Section 4, and Section 5 includes the discussions of the empirical findings. Section 6 is reserved for the conclusion. 2. LITERATURE REVIEW The inequality and the debate on how to solve this problem are vastly argued in the literature. So many different dimensions of this problem have been discussed in the very different researches. The increasing interest on the inequality causing from the redistributive consequences of it, result in an extensive literature on the subject. The consequences of the unequal income distribution reflect the overall economy and the distribution and the redistribution of the income is a matter to individuals. Therefore policy makers give a big attention to this problem. Of course, there will be heterogeneity across the economy however, the extent of this is very important. The inequality of income could be measured at the different levels of economies such as international, continental and national levels. As the economic growth in the economies, especially in the emerging economies, do not distributed equally across the society, this may alter the dispersion between the different income groups and/or different regions. Therefore, one of the dimensions of the interest is focusing on the regional aspect of the inequality. In the literature, the measurement of the inequality and the decomposition of different income factors and population groups have been developed over time. 3

Besides the other dimensions of the inequality, regional inequalities have also been the main object of the extensive researches over the last decade. Researchers focus on the sources of regional dispersion by using different decomposition methods 1. In the literature, so many different countries are examined in the studies by regional distinction (Terrasi, 1999 for Italy; Kim and Jeong, 2003 for Korea; Gluschenko, 2010 for Russia, Brewer and Wren-Lewis, 2012 for Great Britain; Heshmati, 2004 for selected large countries; Mishra and Parikh, 1997 for India, Moffitt and Gottschalk, 2002 for United States; Lee, 2000 and Gustafsson and Shi, 2002 for China, Akita, 2000, Cao and Akita, 2008 for Vietnam; Bouvet, 2007; Cowell and Fiorio, 2010). Country base studies for the different countries do not have any consensus about which subgroup is more effective than the others to overall inequality. These studies reveal whether within-region or between-region inequality is more responsible than the other. Even the growing literature on the regional income inequality for different countries, unfortunately, there is limited studies for Turkey. The researches point out the regional dispersion of the Turkey and show that disparities in the income of the different regions persist its existence throughout years (Doğruel and Doğruel, 2003; Yıldırım and Öcal, 2006; Yıldırım et al, 2009; Filiztekin and Çelik, 2010; Dayıoğlu and Başlevent, 2005, 2006; Karahasan, 2015; Öztürk, 2012). A very new and a comprehensive study on the inequality of the Turkey is a report prepared for Turkish Industry and Business Association (TÜSİAD) in which both household based and regional based inequalities are investigated (Selim et al, 2014). Again, this study the regional inequalities are calculated and the results indicate that the existence of regional inequalities are mainly the contribution to overall inequality is arise from within inequalities in the regions. The Turkish economy has more equal income distribution in the last decade compared to 1990s. There is a gradually improvement in inequality after 2002 2. Even, the improvement in inequality, economists believe that this improvement is insufficient. Because, the comparison of Turkey with other developing countries, the numbers of the inequality of the Turkey reveal that, Turkey still has one of the worst income distribution. The comparison of Turkey with OECD countries and European Union Countries demonstrate that, the overall inequality in the distribution of household 1 In the literature, there exist many different decomposition methods for analyzing of the sources of inequality. One is to decompose overall inequality by income source (Shorrocks, 1982; Jenkins, 1995), the other is to decompose by factor (Fields, 2003) and another one is to decompose by population subgroup (Jenkins, 1995, Mookherjee and Shorrocks, 1982). In this paper, the focus is only on the effects of the population subgroups on the inequality. 2 The empirical analysis of the Turkish households income is a little problematic before the year 2002, due to lack of suitable data set. For sure, there are some researches on this issue; however their results are limited because of the nature of the data set. Only, one comprehensive study is known for that period. Gürsel et al. (2000) prepared a report for TÜSİAD by using the data from the year 1987 and 1994 and investigate the inequality and the sources of the inequality with different dimension. They found that there is a little improvement in the inequality in this period. 4

income is unequal compared to OECD countries and very higher than the European Union Countries. (Figure 1 and 2 about here) The inequality level of the countries which are expressed with Gini is given in the Figures 3. As seen from the tables, Turkey has the worst distribution compared to EU countries and she has only better distribution than Mexico compared to other OECD countries for the year 2012. According to World Bank report, the overall inequality in Turkey is explained by rural-urban differences (World Bank, 2000). Besides, the regional disparities of Turkey are high compared to European Union countries, the disparity ratio of Turkey (the difference between the richest and the poorest NUTS2 regions) is two times higher (5.7%) around than the EU countries (Burrell and Oskam, 2005). 3. METHODOLOGY The income inequality measures are widely utilized in the empirical analysis. Even, there is no specific rule for the choosing the appropriate inequality measures as some of them satisfies some desirable properties, some other measures satisfies the some other ones. Therefore, depending on the main question of the research, the choice of right inequality measure will be changed. The common used, inequality measures in the literature are Gini and The Generalized Entropy measures group. Besides, choosing the right unit of analysis and the choice of the equivalent scale are also important points for the empirical analysis. In this paper, analogously to other empirical studies the household level is taken as a unit of analysis and the inequality measures are calculated by utilizing the overall equivalent individual disposable income 4. 3. Gini coefficient is one of the methods of measuring inequality which will be described more detail in the next section. 4 The dataset of the Turkish surveys cover both household and individual level datas. Even, empirical analysis of the income inequality rely on the individual level incomes, in some of the households, there may exists some individuals with no income, however these individuals will also benefit from the total income of the household which is earned by the other members of the households. Therefore, with assuming the equal sharing of income within a household, an equivalent disposable income measure is created in order to assess the individual based disposable income. For this purpose, the equivalent scale is calculated as e follows: N = S 0 e 1 where S is the household size, e is the elasticity of the scale rate with respect to household size. The elasticity of scale is taken as a value of 0.5 because in the literature, studies mostly used the elasticity of scale as 0,5 (OECD, 1998 and Atkinson; 1995) and in order to make comparable this study with the others, the same value is used in the empirical investigation. At last, The disposable income for the individuals is calculated as follows: Y = R S where Ri and Yij is household total disposable ij i income and individual equivalent disposable income (where i refers to households and j refers individuals). 5 e

Inequality Measures The empirical studies of income inequality use very different inequality measures 5. Gini coefficient is one of the most used one which is a single number index of inequality. It is ranging from 0 to 1 (If the income distribution is completely equal (unequal), the Gini index is equal to zero (one)) and could be expressed as follows: Gini= y y (1) where n is the number of individuals (equivalent households) in the sample, y i and y j are the income of individuals (equivalent households) (1,2,3,, ) and y is the arithmetic mean income 6. Generalised Entropy (I α) class measures are also widely used to express the inequality of the population. The general formula of the members of generalized entropy class of measures is as follows: = ( ) 1 (2) where the parameter α is represents the weight given to distances between incomes at different parts of the income distribution 7. The name of the entropy class measures differ with the different real values among 0, 1 and 2 in practice 8. Decomposing by Subgroups Many factors could contribute the overall inequality. The important question is which one the factors have more importance than the others for the inequality. Therefore, the dispersion of inequality measures with respect to individual characteristics such as gender, age, education level and household characteristics such as number of individual in the household, type of the household and the geographic characteristic such as regions, provinces are interested in. The decomposing the causes of the inequality is very crucial for the policy makers as revealing the importance of the different factors 5 For more information about the inequality measures and their features, see Litchfield, 1999. 6 Researchers discussed deeply about the robustness of all inequality measures. The debates on the results of is rely on the fact that this index is more sensitive to income transfers among middle income groups which means transfers from and to upper and lower tail of the distribution have less effect on the value of the index. However, even this is the fact, Gini index is widely used and therefore in this study, it is also calculated. 7 At the lower α values, the inequality measure is more sensible for the income transfers in the lower tail of the distribution which means this inequality measure is used to give more weights to changes in the income of households located at the lower the lower tail. At the higher α values, the inequality measure is to be more sensitive to changes in the upper tail (Litchfield, 1999). 8 If α=0, the Generalised Entropy measure is known as Theil s L index or the mean log deviation (MLD) measure, If α=1, the measures is called as Theil s T index. At last, if α=2 then, the measure is named as one half the squared coefficient of variation. 6

will improve the effectiveness of the macroeconomic policies. With the disaggregation of the inequality, the particular factor of the inequality could be targeted. Hence, the decomposing by population subgroups is one of the prominent methods for the inequality researches. Regional income inequality arise distributive problems for the society. In order to solve the regional dispersion, it is necessary to identify the contribution of the regions to the inequality. In the basis of this reality, in this study, the overall inequality decomposes into between regions and within regions inequality. For further investigation, regional inequalities are decomposed into different subgroups such as education and gender. Jenkins (1995) provides an additive decomposition method for the population subgroups. He decomposes the total inequality into non overlapping population subgroups and declared the total inequality equals the sum of two contributions, one is within-group and the other is between-group components. He preferred to use the generalized entropy measures as they have the most desirable decomposability properties. The general expression for the sum of the inequalities within each group and the inequality exists between the groups is written as follows: = + In the equation, I represents the within group inequality which is the weighted sum of the inequalities within each subgroup and I represents the between group inequality where the inequality is each individuals income receive the mean income of the subgroup to which this person belonged. Jenkins (1995) shows two different inequality measures for this decomposition, the mean logarithmic deviation (MLD) and half the squared coefficient of variation. The formulas of these inequality measures are as follows, respectively: =(1 ) ( ) (3) = ( ) (4) where n is the number of households with a mean of individual in income category of i. and Y i is an income of The rewriting the equations with respect to subgroup decomposition are as follows: = + (1 ) (5) = ( ) + ( ) 1 (6) 7

Where represents the population share of group k, represents group k s mean income relative to the population mean, group k s share of total population income. The first term of the equation is the within group inequality and whereas the second term of the equation is the between group inequality. In the present paper, one half the squared coefficient of variation, CV is chose for the investigation. This measurement is commonly be employed by the empirical studies which examine the distributional problems in developing countries as this inequality measure gives proportionately more weight to gaps in the upper tail of distribution in measuring income inequality. Therefore, adaption of this measure for Turkey in which higher income gaps among households at the higher income group will be suitable. 4. DATA AND EMPIRICAL RESULTS The availability of the household surveys has been improved for Turkey after 2002 with annul Household Budget Surveys (HBS) and then after the year 2006 more comprehensive data set Survey of Income and Living Conditions (SILC) is applied annually. For this research, Income and Living Conditions Survey conducted by Turkish Statistical Institute (TurkStat) for the years of 2006, 2010 and 2013 is utilized 9. The one of the main aim of the paper is to explore the regional inequalities of Turkey, SILC data set is chosen for the analyze, since this data set covers urban and rural areas and statistical regions (SR) NUTS Level 1, NUTS level 2 and NUTS level 3 within the compliance with European Union 10. At first, brief descriptive statistics and inequality measures are given in this section. Afterwards, the overall inequality is decomposed into population subgroup by NUTS1 regions. Therefore, the within and between subgroup decomposition of the overall inequality will be explored. Lastly, the regional inequalities for each different region will be decomposed into one more subgroup such as education attainment of the household head and gender of the household head. Empirical Results Table 1 informs the brief descriptive summary of households for the different regions for the investigated years. According to summary statistics, the overall household size is vary from 10000 to 15000 and the average household size of the Turkey is around four, 9 In the SILC, the entire of the all settlements within the borders of the Republic of Turkey were included within the scope/sample selection. However, the population in the aged home, elderly house, prisons, military barracks, private hospitals, hotels and child care centers together with the immigrant population were excluded out of the scope (SILC, 2011). 10 In this research, NUTS1 level is chosen for the investigation. NUTS1 is composed of 12 different level of Turkey. Istanbul, West Marmara, Aegean, East Marmara, West Anatolia, Mediterrenean, Central Anatolia, West Black Sea, East Black Sea, Northeast Anatolia, Middleeast Anatolia, Southeast Anatolia. 8

there is a slight decrease in the number throughout years. The mean annual income per household has an increasing trend; it is around 14000 Tl and reaches to 27000 in 2013. The same is true for the mean equivalent incomes of the household. (Table 1 about here) The regional based descriptive statistics reveal that, Istanbul, Aegean and the Mediterranean regions have the highest number of household compared to other regions. The lowest household size is at the East Black sea region. All the household size increased from one year to another for the investigated period. The mean household size is higher at the eastern part (North Eastern Anatolia, central Eastern Anatolia and the South Eastern Anatolia) of the Turkey. This means, households are more crowded in these region compared to western part of the Turkey. The size is around 5-5.5 while it is around 3.5-3.9 at the western part of the Turkey. The general inequality and poverty measures for the investigated years show that, there exists an improvement for the whole sample. The finding about the poverty is shown as the head-count ratio 11. The results of income inequality measure of Gini coefficient appear that the year 2006 has more unequal income distribution than the year 2011. As observed from the table, income inequality for overall economy is decreasing over the time, which means that the income is shared more equally in households (0.42 in 2006 and 0.39 in 2011) over time. The mean annual household income for the all regions increased over time. In 2013, the number is two times higher than it is in 2006. When the regions are compared with each other in the same year, it is explored that lowest mean of income of the households is at the South Eastern Anatolia. North Eastern Anatolia has the second lowest mean annual household income. These results are true for all years. The highest number is at the Istanbul region for the all years. After Istanbul region, East Marmara, West Anatolia and Aegean have the higher numbers than the other regions. Actually, the observed results indicate that there is a dispersion between the different part of the country. At the same year, one region (Istanbul region) nearly has two times higher income than the other one (South Eastern Anatolia). The disparity in the incomes of the households in the regions may deteriorate the overall inequality. Besides, the different levels of income create unequal social and living conditions for the individuals which lead to not only regional income inequality but also regional inequality between social indicators. 11 This ratio is the simplest way of measuring poverty and shows the proportion of the population whose income level is lower than the pre-determined poverty line. However, this index does not show the severity of the poor s. Even the ratio improved over the years, the depth of the poor could worsen for the country. 9

The regional income inequality results are given in the Table 2. The measures of Gini coefficient and half squared coefficient of variation are utilized to reveal the inequality levels. The results of the Gini coefficient point out that some parts of Turkey are more unequal than the other parts. (Table 2 and Figure 3 about here) Istanbul, West Marmara and Central Anatolia region have the lowest Gini coefficient which means the income distribution in these regions is more equal than the regions. The reason of this relatively equal distribution can be a result of the labor market conditions. As, most of the individuals are wage earner in this region, rather than to be a self-employed in the rural area, the individuals could get more equal wages compared to the other parts of the Turkey. On the contrary, Mediterranean and South Eastern Anatolia region have the highest value of Gini coefficient. The value of Gini coefficient in the Mediterranean region is mainly due to the immigration rate of the Mediterranean region. As the low job opportunities in the eastern part of the Turkey force individuals to seek for a job at the nearby regions, Mediterranean region has the highest ratio of immigrants. Besides, the seasonality structure of the jobs may deteriorate the income distribution. The value of Gini coefficient for the South Eastern Anatolia region is also relevant with the expectations. Also, the value of Gini coefficient for the central eastern and south eastern Anatolia is not far from the south eastern Anatolia. As the eastern part of the Turkey is less developed than the other parts, the low productivity causes more inequality in these regions. Decomposition by Subgroups The influence of the population subgroups to inequality is investigated by decomposing the overall inequality. The root of the Jenkins (1995) method is followed in order to achieve the effects of subgroups onto inequality. With decomposition analyze, within different groups and between different groups inequalities explored. So as one of the aim of this paper is to reveal the effect of regions on the overall inequality, we utilized the NUTS1 level of regional boundaries. (Table 3 about here) Table 3 represents the decomposition of the overall inequality into two components: within and between inequality. The decomposition by regions is investigated by using one of the generalized entropy measure, namely half squared of coefficient variation ( ). The exact decomposability of property of ( ) allows the sum of the within and between components to equal to the total inequality. As seen from the table 3, it is obvious that, the within regional inequality is responsible from the overall inequality. 10

The effect of between regional inequality is very low compared to the within inequality. Therefore, one can conclude that, when the inequality in Turkey is decomposed into subgroups by regions, the within inequality in the regions has more effective role on the overall inequality than the inequality across the regions. Besides, the numbers in the table also indicate that the total inequality is improved throughout years. Although there is no difference between the years in terms of between regional inequality, the within regional inequality is improved. Therefore, it is very clear that the improvement in the overall inequality is mainly result from the improvement in the within regional inequality. A further decomposition analyze is done as to reveal the role of two important subgroups role on the regional based inequality. Therefore, the regional inequalities are separately decomposed into two subgroups: educational attainment and gender of the household head. Education consists of six different categories (illiterate, literate, primary school, secondary school, high school, vocational high school, university). Table 4represents the results of decomposition of regional inequalities by educational attainment. Table also shows the inequality of each education categories in the different regions. The inequality of the each subgroup is calculated by using half squared coefficient of variation ( ) and Gini coefficient. (Table 4 about here) It could be stated for all different regions that, the within educational inequality is mainly accounted for the overall regional inequality. In particular, the difference between the inequality within educational subgroups and the inequality between the educational subgroups is very high for all investigated years. When the different years compared to each other, from 2006 to 2010 the within group inequality is deteriorated only for the regions Mediterranean and Central Anatolia. For the other regions, there is an improvement in the inequality within the educational subgroups and this reflects to overall inequality, as well. Although, the effect of between inequality of educational subgroup is very low compared to within inequality, the same pattern is also true for the inequality between the educational subgroups. The inequality measures of the each educational subgroup displays that mainly illiterate and literate subgroups of education have the lowest within inequality for the investigated period and for the all regions. If the regions are compared to each other based on the educational subgroup it is appear that the higher education the higher inequality measure for the within inequality. This result is valid for all regions. The highest value of the inequality measures are seen at the Mediterranean region for the year 2006 and Central Anatolia for the year 2010. 11

The same exercise is made for the different population subgroup which is gender of the household head. Decomposition each region inequality separately both in the within inequality and between inequality components and the inequality measures of the subgroups are given in the Table 5. (Table 5 about here) Table 5 explores the importance of the inequality within the female and male household head for the overall regional inequality. For the each regional inequality, the within inequality of subgroup is headed to overall inequality. The inequality between the subgroups has nearly no importance for all different regions. A striking but not surprising point observed from the table is about the values of inequality measures of subgroups. It is revealed that the inequality in the female household head subgroup is higher than the inequality in male household head subgroup for the all regions, separately. The worst inequality of female household head subgroup is seen at the South East Anatolia region. This conclusion indicates the well-being of female household head subgroup is worse than the male headed group. As many of the females are mostly responsible for the housework and mainly work as an unpaid family worker or unskilled worker, their income level is lower than the others. However the dispersion in the income of the females is not result from this. It heavily rely on the fact that, females mostly attach to the labor market when they have a highest education level (university graduation) or they have lowest education level (illiterate/literate). As the educational attainment differentials cause a wide dispersion of the income of females, ultimately this gap end up with high within income inequality for the females. 5. CONCLUSION The growing concern about the distributive effects of the economic growth on the overall economy makes the inequality one the important socio-economic indicator. Precisely, the developing countries suffer from the unequal income distribution in the overall society as the consequences of the growth for the different parts and/or groups of the country will not the same. Therefore, in order to solve this problem, the differentials between the different parts and/or groups have to be narrowed. The main focus of the paper is the regional income inequality of the Turkey. The regional dispersion of the Turkish economy is examined for the selected years in order to reveal the importance of the regions on the overall inequality. The results reveal that overall inequality in Turkey has improved over the years. However, when the different part of the Turkey is examined more detailed, it is observed that the regions show 12

disparities. The most unequal distribution is explored at the southern parts of the Turkey. After revealing the regional dispersion of the Turkey, in order to obtain whether within inequality of regions or the between inequality across the regions is responsible for the overall inequality, the overall inequality is decomposed into regional subgroups. It is captured that for the all investigated years, the within inequality of the regions contribute the overall inequality. This conclusion leads to research to the determinants of the regional inequality. Afterwards, each region decompose into population subgroups such as education attainment and gender of the household head, and the results indicate that, for all regions, even the magnitude of the number is changed, the contribution of the within inequality of subgroups are higher the inequality between the subgroups. 13

REFERENCES Akita, T. (2000), Decomposing Regional Income Inequality using a Two-Stage Nested Theil Decomposition Method, International Development Working Paper Series, No: 2. Bouvet, F. (2007), Dynamics of Regional Income Inequality in Europe and Impact of EU Regional Policy and EMU 4th DG ECFIN Annual Research Conference of Growth and income distribution in an integrated Europe: Does EMU make a difference? Brussels. Brewer, M., and Wren-Lewis, L. (2012), Accounting for Changes in Inequality since 1968: Decomposition analyses for Great Britain, Institute for Social and Economic Research Working Paper Series, No. 2012-17. Burrell, A. M. and Oskam A. J. (2005), Turkey in the European Union: Implications for Agriculture, Food and Structural Policy, CABI publishing. Cao, T. C. V. ve Akita, T. (2008), Urban and Rural Dimensions of Income Inequality in Vietnam, GSIR Working Papers, Economic Development & Policy Series, No: EDP08-2 Cowell, F. A. and Fiorio, C. V. (2010), Inequality Decompositions, Growing Inequalities Impacts Discussion Paper, No: 4. Dayıoğlu, M. and Başlevent, C. (2005), A Household Level Examination of Regional Income Disparity in Turkey, METU Studies in Development, Vol. 32 (2), pages: 275-302. Dayıoğlu, M. and Başlevent, C. (2006), Imputed Rents and Regional Income Inequality in Turkey: A Subgroup Decomposition of the Atkinson Index, Regional Studies, Vol. 40(8), pages: 889-905. Doğruel, F. and Doğruel, S. (2003), Türkiye de Bölgesel Gelir Farklılıkları ve Büyüme, İktisat Üzerine Yazılar I Küresel Düzen: Birikim, Devlet ve Sınıflar Korkut Boratav a Armağan, Eds. A.Köse, F.Şenses, E.Yeldan (in Turkish, book chapter). Filiztekin, A. Çelik, M. A. (2010), Regional Income Inequality in Turkey, Megaron, Vol 5(3), pages: 116-127 Gluschenko, K. (2010), Methodologies of Analyzing Inter-Regional Income Inequality and Their Applications to Russia, William Davidson Institute Working Paper, No. 984. Gustafsson, B. and Shi, L. (2002), Income Inequality Within and Across Counties in Rural China 1988 and 1995, Journal of Development Economics, Vo. 69 (1), pages: 179-204. Gürsel, S., Levent, H., Selim, R. and Sarıca, Ö. (2000). The Household Income Distribution and Poverty in Turkey: Comparison with European Union, TÜSİAD Working Paper, No: TÜSİAD-T/2000-12/295, Ankara (in Turkish). Heshmati, A. (2004), Regional Income Inequality in Selected Large Countries, IZA Discussion Paper Series, No: 1307. Jenkins, S. P. (1995), Accounting for Inequality Trends: Decomposition Analyses for the UK, 1971 86, Economica, Vol. 62 (1), pages: 29-63. Karahasan, B. C. (2015), Regional Inequalities in Turkey: Post 2001 Era, Marmara Universitesi İ. İ. B. Dergisi, Vol. 37 (1), pages: 125-147. Kim, E. and Jeong, Y. H. (2003), Decomposition of Regional Income Inequality in Korea, The Review of Regional Studies, Vol. 33 (3), pages: 313-327. 14

Litchfield, J. A. (1999). Inequality: Methods and Tools, The World Bank, Washington, D.C. Lee, J. (2000), Changes in the sources of China s regional inequality, China Economic Review, Vol. 11 (3), pages: 232-245. Mishra, P. and Parikh, A. (1997), Distributional Inequality in Indian States, Journal of Income Distribution, Vol.7 (1), 89-103. Moffitt, R. A. and Gottschalk, P. (2002), Trends in the transitory variance of earnings in the United States, The Economic Journal, Vol.112. Mookherjee, D. and Shorrocks, A. (1982), "A decomposition analysis of the trend in UK income inequality" The Economic Journal, Vol. 92(368), pages: 886-902. OECD. (2014). Society at a Glance 2011, OECD Books. Öztürk, L. (2012), Türkiye de Bölgesel Eşitsizliğin Sektörel Dinamikleri: Bir Ayrıştırma Analizi, 1987-2001, Ege Akademik Bakış Dergisi, Cillt 12, No: 2. Selim, R., Günçavdı, Ö. and Bayar, A. A. (2014), Household Income Inequality in Turkey: Functional Income Sources and Regional Ineqaulities, TÜSİAD Report, No: TÜSİAD-T/2014-6/554. Shorrocks, A. F. (1982). Inequality Decomposition by Factor Components, Econometrica, Vol. 50, pp. 193-212. Shorrocks, A. F. (1983). The Impact of Income Components on the Distribution of Family Incomes, Quarterly Journal of Economics, Vol. 98, pp. 311-326. Terrasi, M. (1999), Convergence and Divergence across Italian Regions, Annals of Regional Science, Vol. 33 (4), pages: 491-510. World Bank, (2000), World Development Report: Attacking Poverty, New York: Oxford University Press Yıldırım, J. and Öcal, N. (2006), Income Inequality and Economic Convergence in Turkey, Transition Studies Review, Vol.13 (3), pages: 559-568. Yıldırım, J., Öcal, N. and Özyıldırım, S. (2009), Income Inequality and Economic Convergence in Turkey: A Spatial Effect Analysis, International Regional Science Review, Vol. 32 (2), pages: 221-254 15

Table 1 - General Summary of the Samples 2006 2010 Total Sample Size 10915 12102 Median Household Size 4 3 Mean Household size 3,92 3,75 Mean annual income per household 13889 21116 Mean equivalent annual income per household 7638 11822 İstanbul Sample Size 1208 1517 Median Household Size 3 4 Mean Household size 3,47 3,58 Mean annual income per household 19988 29876 Mean equivalent annual income per household 11444 16888 West Marmara Sample Size 873 894 Median Household Size 3 3 Mean Household size 3,15 2,99 Mean annual income per household 12814 18476 Mean equivalent annual income per household 7481 11077 Aegean Sample Size 1468 1750 Median Household Size 3 3 Mean Household size 3,32 3,14 Mean annual income per household 15390 22473 Mean equivalent annual income per household 8981 13372 East Marmara Sample Size 858 923 Median Household Size 3 3 Mean Household size 3,49 3,46 Mean annual income per household 16687 21747 Mean equivalent annual income per household 9461 12268 West Anatolia Sample Size 926 1072 Median Household Size 4 3 Mean Household size 3,70 3,49 Mean annual income per household 17179 24604 Mean equivalent annual income per household 9572 13950 Mediterranean Sample Size 1109 1234 Median Household Size 4 3 Mean Household size 3,74 3,34 Mean annual income per household 11665 20226 Mean equivalent annual income per household 6438 11833 Central Anatolia Sample Size 757 829 Median Household Size 4 4 Mean Household size 3,93 3,78 Mean annual income per household 12971 18917 Mean equivalent annual income per household 6852 10243 West Black sea Sample Size 810 769 Median Household Size 4 3 Mean Household size 3,91 3,48 Mean annual income per household 11728 17773 Mean equivalent annual income per household 6238 10102 East Black Sea Sample Size 669 597 Median Household Size 4 4 Mean Household size 3,61 3,72 Mean annual income per household 14218 18879 Mean equivalent annual income per household 7793 10203 16

Table 1 - General Summary of the Samples, continued 2006 2010 North Eastern Anatolia Sample Size 717 771 Median Household Size 5 4 Mean Household size 5,04 4,70 Mean annual income per household 10736 16973 Mean equivalent annual income per household 5121 8543 Central Eastern Anatolia Sample Size 630 837 Median Household Size 5 5 Mean Household size 5,23 5,12 Mean annual income per household 10752 18532 Mean equivalent annual income per household 5156 9190 South Eastern Anatolia Sample Size 890 909 Median Household Size 5 5 Mean Household size 5,51 5,25 Mean annual income per household 8120 15136 Mean equivalent annual income per household 3747 7306 Source: Author s calculations from the data set of TurkStat for the year 2006, 2010 and 2013. Table 2. Overall and Regional Income Inequality Measures 2006 2010 2013 Gini coefficient İstanbul 0.37 0.36 0.41 West Marmara 0.35 0.36 0.36 Aegean 0.4o 0.38 0.37 East Marmara 0.39 0.33 0.32 West Anatolia 0.40 0.36 0.39 Mediterranean 0.41 0.39 0.40 Central Anatolia 0.34 0.36 0.38 West Black Sea 0.36 0.34 0.37 East Black Sea 0.38 0.32 0.31 North Eastern Anatolia 0.37 0.39 0.36 Central Eastern Anatolia 0.40 0.39 0.37 South Eastern Anatolia 0.39 0.39 0.36 Turkey-Total 0.42 0.39 0.41 Source: Author s calculations from the data set of TurkStat for the year 2006, 2010 and 2013. Table 3: Decomposition of inequality into subgroups: by Regions (using I(2) inequality measure) 2006 2010 2013 Total Inequality 0.48 0.48 0.51 Between Region Inequality 0.04 0.03 0.04 Within Region Inequality 0.44 0.45 0.47 Source: Author s calculations from the data set of TurkStat for the year 2006, 2010 and 2013. 17

Table 4: Decomposition of inequality into subgroups: by educational attainment of the household head 2006 2010 2013 I (2) Gini I (2) Gini I (2) Gini ISTANBUL Illiterate 0,149 0,291 0,259 0,301 0,144 0,224 Literate 0,148 0,299 0,128 0,282 0,134 0,282 Primary 0,176 0,301 0,193 0,303 0,195 0,311 Secondary 0,301 0,357 0,135 0,264 0,115 0,274 High School 0,327 0,367 0,159 0,277 0,149 0,306 Vocational high school 0,217 0,298 0,152 0,281 0,515 0,412 University 0,294 0,375 0,314 0,372 0,371 0,413 Within 0,288 0,287 0,378 Between 0,071 0,100 0,122 WEST MARMARA Illiterate 0,170 0,329 0,181 0,289 0,114 0,267 Literate 0,170 0,296 0,166 0,313 0,123 0,274 Primary 0,211 0,318 0,231 0,320 0,162 0,306 Secondary 0,273 0,339 0,150 0,292 0,076 0,215 High School 0,153 0,271 0,183 0,317 0,242 0,338 Vocational high school 0,188 0,305 0,149 0,287 0,115 0,238 University 0,191 0,306 0,195 0,291 0,250 0,336 Within 0,220 0,220 0,227 Between 0,044 0,066 0,08 AEGEAN Illiterate 0,176 0,307 0,206 0,292 0,108 0,256 Literate 0,370 0,353 0,363 0,320 0,111 0,250 Primary 0,297 0,354 0,207 0,318 0,149 0,298 Secondary 0,370 0,396 0,235 0,296 0,092 0,232 High School 0,278 0,368 0,296 0,353 0,175 0,302 Vocational high school 0,246 0,350 0,289 0,323 0,121 0,279 University 0,350 0,378 0,473 0,363 0,583 0,380 Within 0,377 0,403 0,471 Between 0,100 0,085 0,068 EAST MARMARA Illiterate 0,130 0,243 0,142 0,291 0,140 0,292 Literate 0,362 0,342 0,159 0,282 0,219 0,373 Primary 0,424 0,340 0,214 0,286 0,328 0,328 Secondary 0,356 0,410 0,171 0,302 0,129 0,251 High School 0,346 0,329 0,105 0,247 0,135 0,269 Vocational high school 0,138 0,277 0,556 0,343 0,151 0,301 University 0,337 0,377 0,207 0,310 0,131 0,266 Within 0,388 0,270 0,241 Between 0,070 0,042 0,025 WEST ANATOLIA Illiterate 0,209 0,328 0,141 0,269 0,147 0,217 Literate 0,142 0,275 0,240 0,342 0,137 0,263 Primary 0,241 0,331 0,184 0,310 0,123 0,250 Secondary 0,316 0,351 0,172 0,319 0,225 0,331 High School 0,155 0,291 0,155 0,279 0,187 0,277 Vocational high school 0,442 0,422 0,295 0,327 0,371 0,405 University 0,245 0,348 0,188 0,301 0,164 0,311 Within 0,319 0,223 0,217 Between 0,101 0,070 0,096 18

Table 4: Decomposition of inequality into subgroups: by educational attainment of the household head, continued. 2006 2010 2013 I (2) Gini I (2) Gini I (2) Gini MEDITERRANEAN Illiterate 0.632 0.390 0.115 0.260 0,251 0,389 Literate 0.345 0.414 0.134 0.288 0,105 0,241 Primary 0.373 0.383 0.259 0.328 0,196 0,316 Secondary 0.211 0.298 0.194 0.280 0,243 0,269 High School 0.733 0.450 0.802 0.402 0,350 0,403 Vocational high school 0.241 0.359 0.392 0.346 0,157 0,300 University 0.184 0.284 0.860 0.414 0,261 0,360 Within 0.424 0.702 0.291 Between 0.070 0.108 0.106 CENTRAL ANATOLIA Illiterate 0.093 0.238 0.179 0.296 0,275 0,413 Literate 0.329 0.330 0.186 0.296 0,121 0,225 Primary 0.303 0.346 0.210 0.320 0,210 0,333 Secondary 0.143 0.263 0.383 0.324 0,090 0,238 High School 0.246 0.337 0.489 0.328 0,463 0,490 Vocational high school 0.103 0.234 0.169 0.270 0,085 0,223 University 0.145 0.284 0.587 0.360 0,107 0,220 Within 0.243 0.420 0.253 Between 0.030 0.060 0.098 WEST BLACK SEA Illiterate 0.147 0.304 0.103 0.250 0,212 0,349 Literate 0.219 0.326 0.145 0.283 0,178 0,322 Primary 0.203 0.331 0.216 0.318 0,178 0,321 Secondary 0.404 0.345 0.198 0.302 0,877 0,397 High School 0.202 0.311 0.115 0.231 0,442 0,354 Vocational high school 0.154 0.309 0.170 0.304 0,130 0,212 University 0.494 0.326 0.154 0.280 0,171 0,310 Within 0.337 0.193 0.304 Between 0.060 0.053 0.057 EAST BLACK SEA Illiterate 0.227 0.321 0.072 0.211 0,298 0,385 Literate 0.100 0.256 0.127 0.284 0,340 0,320 Primary 0.265 0.339 0.194 0.309 0,115 0,252 Secondary 1.044 0.475 0.193 0.301 0,000 0,000 High School 0.259 0.331 0.221 0.285 0,013 0,079 Vocational high school 0.452 0.313 0.100 0.252 0,007 0,051 University 0.199 0.288 0.145 0.280 0,083 0,218 Within 0.422 0.179 0.109 Between 0.042 0.030 0.070 NORTH EAST ANATOLIA Illiterate 0.132 0.264 0.207 0.332 0,051 0,138 Literate 0.357 0.391 0.153 0.272 0,422 0,477 Primary 0.234 0.347 0.304 0.350 0,287 0,415 Secondary 0.166 0.296 0.226 0.329 0,334 0,437 High School 0.214 0.326 0.117 0.244 0,121 0,234 Vocational high school 0.133 0.277 0.175 0.318 0,073 0,217 University 0.241 0.236 0.226 0.340 0,036 0,148 Within 0.240 0.274 0.175 Between 0.070 0.100 0.034 19

Table 4: Decomposition of inequality into subgroups: by educational attainment of the household head, continued. 2006 2010 2013 I (2) Gini I (2) Gini I (2) Gini CENTRAL EAST ANATOLIA Illiterate 0.318 0.365 0.139 0.270 0,340 0,384 Literate 0.276 0.372 0.152 0.277 0,099 0,218 Primary 0.307 0.381 0.193 0.317 0,523 0,418 Secondary 0.195 0.315 0.321 0.340 0,151 0,306 High School 0.222 0.326 0.343 0.338 0,039 0,154 Vocational high school 0.065 0.201 0.180 0.313 0,166 0,314 University 0.423 0.372 0.272 0.327 0,287 0,379 Within 0.357 0.322 0.283 Between 0.074 0.145 0.037 SOUTH EAST ANATOLIA Illiterate 0.225 0.340 0.213 0.318 0,207 0,358 Literate 0.482 0.383 0.194 0.305 0,143 0,300 Primary 0.335 0.363 0.413 0.360 0,145 0,272 Secondary 0.351 0.385 0.201 0.320 0,069 0,188 High School 0.261 0.352 0.174 0.303 0,085 0,198 Vocational high school 0.307 0.362 0.175 0.305 0,130 0,260 University 0.160 0.284 0.550 0.352 0,305 0,344 Within 0.334 0.433 0.218 Between 0.050 0.092 0.218 Source: Author s calculations from the data set of TurkStat for the year 2006, 2010 and 2013. 20

Table 5. Decomposition of inequality into subgroups: by gender of the household head 2006 2010 2013 I (2) Gini Gini Gini I (2) I (2) ISTANBUL Female 0.313 0.379 0.256 0.347 0,578 0,462 Male 0.362 0.365 0.406 0.363 0,494 0,408 Within 0.357 0.388 0.500 Between 0.001 0.000 0.000 WEST MARMARA Female 0.297 0.374 0.195 0.311 0,180 0,326 Male 0.260 0.343 0.295 0.361 0,310 0,363 Within 0.262 0.286 0.306 Between 0.001 0.000 0.001 AEGEAN Female 0.244 0.3677 0.379 0.377 0,256 0,393 Male 0.499 0.41661 0.506 0.382 0,553 0,372 Within 0.471 0.489 0.538 Between 0.000 0.000 0.001 EAST MARMARA Female 0.396 0.365 0.239 0.336 0,108 0,265 Male 0.456 0.386 0.320 0.326 0,286 0,328 Within 0.454 0.312 0.264 Between 0.000 0.001 0.002 WEST ANATOLIA Female 0.364 0.405 0.273 0.368 0,396 0,425 Male 0.426 0.40367 0.296 0.357 0,309 0,388 Within 0.420 0.293 0.312 Between 0.000 0.001 0.001 MEDITERRANEAN Female 0.739 0.448 0.276 0.322 0,339 0,396 Male 0.466 0.409 0.861 0.398 0,383 0,395 Within 0.491 0.807 0.387 Between 0.000 0.002 0.010 CENTRAL ANATOLIA Female 0.226 0.303 0.235 0.341 0,235 0,360 Male 0.267 0.339 0.489 0.356 0,341 0,383 Within 0.265 0.478 0.345 Between 0.001 0.002 0.007 WEST BLACK SEA Female 0.183 0.329 0.279 0.316 0,146 0,307 Male 0.401 0.362 0.244 0.340 0,367 0,377 Within 0.392 0.247 0.355 Between 0.002 0.000 0.006 EAST BLACK SEA Female 0.413 0.397 0.139 0.290 0,293 0,420 Male 0.465 0.373 0.215 0.318 0,175 0,306 Within 0.462 0.209 0.179 Between 0.000 0.001 0.001 21

Table 5: Decomposition of inequality into subgroups: by gender of the household head 2006 2010 2013 I (2) Gini Gini I (2) I (2) Gini NORTH EAST ANATOLIA Female 0.271 0.346 0.203 0.330 0,279 0,405 Male 0.303 0.368 0.385 0.389 0,205 0,355 Within 0.301 0.373 0.207 Between 0.001 0.001 0.002 CENTRAL EAST ANATOLIA Female 0.177 0.312 0.208 0.313 0,690 0,498 Male 0.443 0.399 0.475 0.394 0,237 0,353 Within 0.431 0.466 0.311 Between 0.001 0.001 0.009 SOUTH EAST ANATOLIA Female 0.305 0.389 0.320 0.402 0,128 0,275 Male 0.376 0.386 0.545 0.383 0,332 0,362 Within 0.373 0.525 0.327 Between 0.001 0.001 0.000 Figure 1: Gini of OECD countries for the year 2012. 0,48 0,44 0,4 0,36 0,32 0,28 0,24 0,2 0,16 0,12 0,08 0,04 0 Source: OECD Social and Welfate Database, 2012. 22

Figure 2: Gini of EU countries and Turkey for the year 2012. 0,4 0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 Source: OECD Social and Welfare Database, 2012. Figure 3: Gini s of Regions of the Turkey for the years 2006, 2010 and 2013. 0,45 0,43 0,41 0,39 0,37 0,35 0,33 0,31 0,29 0,27 0,25 2006 2010 2013 İstanbul West Marmara Aegean East Marmara West Anatolia Mediterrian Central Anatolia West Black Sea East Black Sea North East Anatolia Central East Anatolia South East Anatolia Turkey-Total Source: Author s calculations from the data set of TurkStat for the year 2006, 2010 and 2013. 23