Technical Report 262- Income Inequality in Greece

7/31/2019 Technical Report 262- Income Inequality in Greece

1/37

1

Income Inequality Measurement in Greece

and Alternative Data Sources: 1957-2009

K. Chrissis, A. Livada

Athens University of Economics and Business

Technical Report No 262

ATHENS UNIVERSITY OF ECONOMICS AND BUSINESS

DEPARTMENT OF STATISTICS

OCTOBER 2012


2/37

2


3/37

3

Abstract

The main objective of this discussion paper is the estimation of income inequality in

Greece for the period 1957-2009. Alternative income sources are used for the

estimation of aggregate and disaggregate measures. Empirical evidence from

tabulated tax data indicates an increase on aggregate income inequality. This view is

not supported by estimates derived from other data sources (i.e. Household

Expenditure Survey). The level of aggregate inequality, also, differs from other

empirical results. These findings imply that different data sources and/or

methodological approaches could lead to different conclusions for the direction and/or

level of aggregate income inequality. Nevertheless, top income shares yield similar

trend (for certain periods) and level (to the possible extend) regardless the data

sources. This view is consistent with Leigh (2007) that top income shares may be a

useful substitute for other measures of inequality.


4/37

4


5/37

5

Income Inequality Measurement in Greece and Alternative Data

Sources: 1957-2009

1. Introduction

This paper provides empirical evidence for income inequality in Greece. Alternative

data sources and methodologies are applied and inequality measures are provided.

More specifically, in section 2 empirical time-series evidence on economic inequality

from grouped tax data will be presented. The time period of the analysis is from the

year 1957 to the year 2009. Micro data are utilized in section 3; data are from EU

SILC for the period 2002-2009. In all cases corresponding evidence from other

countries are presented. In section 4 empirical results from other studies utilizing

other sources [European Community Household Panel (ECHP) and HouseholdExpenditure Survey (HES) micro data] are discussed. Section 5 displays the

comparison for all results of aggregate income inequality. The summary of the

empirical findings are presented in the section 6.

2. Aggregate measures of income inequality from grouped tax data

2.1. Estimation of aggregate measures of income inequality from grouped tax

data

Tax data provide detailed information on nominal family income and its sources, as

reported annually in tax declaration forms. Family income is the sum of income

received by the husband and/or wife. This definition also includes single persons.

These data are compiled by the Tax Authorities and have been published annually by

the National Statistical Service of Greece (NSGG, now ELSTAT) since 1958. From

2003 onwards the publication is conducted by General Secretariat of Informatics

Systems of Ministry of Finance.

Total family income is the sum of one or more of the following components1

a) Income from employmentb) Income from buildings and lease of landc) Income from securitiesd) Income from commercial and industrial enterprisese) Income from agricultural enterprisesf) Income from self-employmentg) Income from abroad

1The category (b) is a merging of two categories after 1984 (economic year 1985): Income from

buildings and Income from lease of land excluding buildings. Also category (d) is divided in two sub-

categories for the years 1976-1998 according to the type of the enterprise. Moreover, imputedincome was introduced in 1997 and its impact is in low levels: around 3% for the years 1997-2002 and

around 0,5% for the rest years.


6/37

6

The tax declarations are submitted in the following year of the year of reference. The

term economic year t refers to income that was acquired in the previous year. Thus,

economic year 2010 refers to the calendar year 2009. Tax data are reported in

tabulated form (grouped tax data). During the whole period the number of classes haschanged, being more analytical in the latter years.

Taking into consideration the issues addressed for the data consistency in the previous

section, the summary inequality measures are compiled. The following indices have

been estimated for the declared income of the physical persons (grouped tax data).

Gini Coefficient (G) Relative Mean Deviation (M) Atkinson Index () (=0,5) Atkinson Index () (=1,5)

General Entropy (GE(0)

TheilsL

or Mean Log Deviation) (a=0) General Entropy (GE(1)Theils T) (a=1) General Entropy (GE(2) type of Coefficient of Variation2- CV) (a=2)

The choice of these indices is based on the underlying properties. Furthermore, these

aggregate indices are widely used for the empirical measurement of inequality.

The distribution of the data within each class is not known. This issue is being tackled

using interpolation methods (Cowell and Mehta, 1982). Two interpolation methods

were used: the split-histogram interpolation method and the linear interpolation

method. The mean value of the computation of these two techniques provides the final

estimation of the measure. The lower and upper bounds of the estimation have been

also compiled. The compiled index of Relative Mean Deviation refers only to lower

bound.

The estimates for the aggregate inequality measures are presented in Table 1. The

following charts are the graphical illustration of the time series of each individual

index.

2The monotonic transformation is GE(2)=(CV^2)/2


7/37

7

Table 1. INCOME INEQUALITY MEASURES

ATKINSON 0,5 ATKINSON 1,5GENERAL

ENTROPY 0

GENERAL

ENTROPY 1

GENERAL

ENTROPY 2GINI

RELATIVE

MEAN

DEVIATION

1957 0,150620 0,313287 0,282220 0,392954 1,287735 0,413949 0,591613

1958 0,133402 0,293722 0,254428 0,331153 0,787963 0,393548 0,571844

1959 0,130009 0,286246 0,246837 0,325014 0,827603 0,387022 0,559498

1960 0,137087 0,303968 0,264442 0,337326 0,820693 0,401974 0,579896

1961 0,139420 0,312264 0,271550 0,339821 0,794060 0,407225 0,590676

1962 0,120729 0,273980 0,232391 0,291718 0,625213 0,376434 0,536807

1963 0,125339 0,284365 0,242405 0,303249 0,676942 0,384327 0,552220

1964 0,124811 0,284268 0,241968 0,300364 0,667374 0,384666 0,553312

1965 0,139982 0,336113 0,281239 0,332894 0,776530 0,408418 0,582500

1966 0,143293 0,343765 0,289680 0,338545 0,749922 0,414897 0,595419

1967 0,129247 0,324962 0,262817 0,300990 0,599381 0,391092 0,558320

1968 0,125803 0,320336 0,256601 0,291591 0,630771 0,385709 0,550536

1969 0,144667 0,389799 0,310289 0,327026 0,620345 0,413822 0,586860

1970 0,142744 0,388022 0,307042 0,320998 0,617437 0,411215 0,584588

1971 0,139517 0,381534 0,299139 0,314016 0,598081 0,406300 0,576680

1972 0,143147 0,389228 0,307514 0,322473 0,596719 0,411731 0,577988

1973 0,150104 0,394428 0,316698 0,365569 1,183395 0,417382 0,555033

1974 0,142381 0,386605 0,305601 0,325717 0,910893 0,408982 0,535146

1975 0,136996 0,387459 0,301029 0,299029 0,535721 0,403768 0,559172

1976 0,137011 0,394175 0,304480 0,294920 0,474531 0,404392 0,575053

1977 0,134153 0,388730 0,298773 0,287544 0,530121 0,399863 0,554751

1978 0,142898 0,483992 0,342084 0,296211 0,433088 0,407658 0,564951

1979 0,145295 0,512957 0,356077 0,298338 0,439562 0,408902 0,580491

1980 0,147218 0,536603 0,367969 0,298940 0,431325 0,410685 0,575034

1981 0,154745 0,589858 0,400694 0,313500 0,519844 0,417930 0,594532

19820,149913 0,617441 0,404044 0,291436 0,379982 0,408140 0,578759

1983 0,151827 0,643488 0,416844 0,294052 0,415320 0,408899 0,577778

1984 0,152509 0,665556 0,426127 0,291825 0,367453 0,409391 0,573240

1985 0,145880 0,677322 0,410515 0,282211 0,449396 0,398079 0,561607

1986 0,146047 0,693310 0,414226 0,281261 0,373141 0,398971 0,562811

1987 0,150449 0,704507 0,430103 0,288476 0,377859 0,404288 0,572324

1988 0,146861 0,668302 0,404701 0,285277 0,368121 0,403161 0,549920

1989 0,156314 0,706447 0,439141 0,301635 0,379882 0,415713 0,531819

1990 0,160726 0,764077 0,467887 0,308334 0,382023 0,420293 0,584255

1991 0,171794 0,848538 0,532496 0,328888 0,417591 0,432737 0,615064

1992 0,169085 0,625355 0,453717 0,329657 0,417436 0,434447 0,617577

1993 0,170262 0,629912 0,456935 0,333068 0,425086 0,436485 0,619126

1994 0,198601 0,739465 0,580770 0,378535 0,476077 0,463643 0,663650

1995

0,198505 0,736807 0,576938 0,380320 0,482901 0,464903 0,663274

1996 0,212691 0,770318 0,631831 0,408173 0,532044 0,480464 0,690938

1997 0,214098 0,773032 0,633260 0,414321 0,555770 0,483906 0,696578

1998 0,216589 0,772651 0,635514 0,423490 0,586885 0,488239 0,704996

1999 0,224316 0,765948 0,641472 0,452053 0,701494 0,500150 0,722323

2000 0,220672 0,770963 0,633222 0,445673 0,700373 0,495808 0,714484

2001 0,219531 0,848531 0,676182 0,438636 0,682915 0,490765 0,706877

2002 0,221107 0,853551 0,680800 0,444225 0,811646 0,493436 0,711761

2003 0,221719 0,871791 0,701416 0,442247 0,707039 0,491397 0,706715

2004 0,233857 0,915198 0,801579 0,454708 0,705327 0,498817 0,717045

2005 0,220999 0,863092 0,700273 0,437325 0,863524 0,488780 0,701187

2006 0,218191 0,863773 0,691441 0,431890 0,716971 0,485528 0,695863

2007 0,216697 0,871003 0,693105 0,428522 0,680229 0,483031 0,691951

2008 0,221513 0,879645 0,714270 0,437052 0,681033 0,489687 0,703828

2009 0,230772 0,898989 0,768572 0,449830 0,669007 0,498291 0,717089

Source: Authors calculations


8/37

8

Looking the period as a whole the empirical results are summarized as follows.

The Gini coefficient implies an increase of inequality. It arises from 0,413949 in 1957

to 0,498291 in 2009. The upward trend seems to take place from the early 1990s,

being relatively steady in the previous period.

The Relative Mean Deviation suggests an increase as well, starting with a value of0,591613 in 1957 and reaching the level of 0,717089 in 2009. The upward trend, as in

the case of Gini, emerges from the early 1990s.

The two variations of Atkinson index yield the same results, indicating that is, an

increase in income inequality. Atkinson and are 0,150620 and 0,313287

respectively for the year 1957 and 0,230772 and 0,898989 respectively for the year

2009.

The Mean Log Deviation (GE(0)) implies an increase of income inequality, with

values of 0,282220 and 0,768572 for the years 1957 and 2009.

The Theils Index (GE(1)) suggests, also, an increasing trend of inequality during the

reference period. It starts at 0,392954 in 1957 and reach the level of 0,449830 in

2009.

The monotonic transformation of Coefficient of Variation (GE(2)) suggests a decline

of inequality during the period, although the trend is not steady. It, also presents cases

of outliers, especially for years 1957, 1973 and 19743.

According to the empirical findings, six indices indicate an increase of income

inequality while one (GE (2)) indicates the opposite (decrease).

The mathematical results are similar to previous studies [Livada (1988) (1991),

Livada and Tsakloglou (1993), Dimelis and Livada (1994) (1997)], but the

conclusions differ due to the quite different reference period.


3Adjustments have made for six years.

0.00

0.05

0.10

0.15

0.20

0.25

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 1-ATKINSON 0,5


9/37

9





0.00

0.10

0.20

0.30

0.400.50

0.60

0.70

0.80

0.90

1.00

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 2-ATKINSON 1,5

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 3-GENERAL ENTROPY 0 - N (MLD)

0.00

0.05

0.10

0.15

0.20

0.25

0.30

0.35

0.40

0.45

0.50

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 4-GENERAL ENTROPY 1 - THEIL T

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 5-GENERAL ENTROPY 2-CV


10/37

10


Source: Author calculations

0.00

0.10

0.20

0.30

0.40

0.50

0.60

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 6-GINI- G

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 7-RELATIVE MEAN DEVIATION-M


11/37

11

2.2. Time series analysis

The following table illustrates the results for Least Square linear regression corrected

with AR(1)4

for the time series of the estimated inequality measures. In all cases

except General Entropy (2) a positive trend appears; the slope coefficient is

statistically significant at 1%. The GE(2) seems to be better described by the quadraticmodel according to Table 3.

Table 2. Trend for aggregate inequality measures

MEASURES CONSTANT COEFFICIENT AR(1) R-SQ

Atkinson_0,5 0,088968

(5,572799)

(0,002656)

(6,123914)

0,810452

(11,89312)

0,955844

Atkinson_1,5 0,195963

(5,952647)

0,013994

(13,99147)

0,613232

(5,649480

0,964081

GE_0 (Mean Log Dev) 0,115026

(2,999479)

0,011743

(10,37721)

0,710967

(7,793125)

0,964305

GE_1 (Theil T) 0,203592

(3,980095)

0,004553

(3,432855)

0,839840

(14,13672)

0,904415

GE_2 (CV) 0,549742

(4,984390)

0,000987*

(0,298558)

0,633000

(6,956647)

0,510394

Gini 0,348991

(16,95194)

0,002803

(4,982442)

0,814197

(11,29531)

0,936205

Rel Mean Deviation 0,480551

(10,92479)

0,004398

(3,728127)

0,836165

(11,85611)

0,914274


Note: All values significant at 1% level (except where *) - * Not statistically significant

Table 3. Quadratic model for GE_2 (CV)MEASURES CONSTANT COEFFICIENT COEFFICIENT 2 AR (1) R-SQ

GE_2 (CV) 0,889414

(8,535418)

-0,028251

(-3,369575)

0,000488

(3,399722)

0,416900

(3,393311)

0,565948


Note: All values significant at 1% level

The summary tests (p-values) ofJarque Bera (testing whether the series of residuals

is normally distributed H0: normal distribution), Durbin-Watson and Godfrey-

Breusch (testing for autocorrelation and serial correlation respectively: DW-d

around 2 no autocorrelation and H0: no serial correlation) and White5

(testing for

heteroscedasticityH0: no heteroscedasticity) are presented in the following table.

The sign of the coefficient, its p-value and the r-square are also included

Table 4. Consistency tests for time series models

1957-2009 Sign p-value R-square Normality DW-d LM(-2) White

Atk 0,5 + 0,00 0,96 0,13 2,04 0,70 0,73

Atk 1,5 + 0,00 0,96 0,00 1,94 0,27 0,23GE_0_mld + 0,00 0,96 0,00 2,20 0,18 0,11GE_1_Theil + 0,00 0,91 0,61 1,91 0,60 0,02GE_2_CV_sq -

+0,000,00

0,57 0,00 1,63 0,15 0,79

GINI + 0,00 0,94 0,37 1,94 0,36 0,04RMD + 0,00 0,91 0,86 1,98 0,20 0,18Source: Authors calculations

4OLS indicated strong evidence of autocorrelation

5This is the original White test, that is with cross terms


12/37

12

All models have statistically significant coefficients except GE(2). In all cases the

trend yield a positive sign. The regression that seems to fit for GE(2) is presented in

Table 3; coefficients are statistically significant in 1% level with OLS AR(1) of

second order. Table 4 summarizes the results for the consistency of the models.

Most of the estimations (Atkinson_0,5, GE(1), Gini and RMD) succeed in thenormality of residuals test. The significant issue is that with the insertion of correction

term AR(1) the problem of autocorrelation is resolved. Moreover, the models do not

face issues of heteroscedasticity (only for 1% for GE(1) and Gini). Finally, taking into

account the high values of the r-square, the models are satisfactory enough for a

typical time series analysis.

In conclusion, it seems that all summary inequality measures, except GE(2), indicate

an upward trend for the period 1957-2009, whereas GE(2) indicate a decline followed

by an increase (explaining thus the quadratic model of description). Nevertheless, the

value of GE(2) never reached its initial level.

2.3. Business cycles characteristics

In this section the business cycle characteristics of the estimated aggregate income

inequality measures are calculated as supplementary evidence to the typical time

series analysis conducted in the previous section. The smoothed trend is estimated

according to Hodrick and Prescott (1980) methodology. The Hodrick and Prescott

methodology (HP filter) is a popular detrending technique due to the flexibility,

simplicity and well-defined criteria on which it was designed. Moreover, the

frequency power rule of Ravn and Uhlig (2002) is applied6

(the figure of powersuggested is 4 meaning that is 6,25). The methodology has as follows. For each

series the trend component is estimated using the HP filter (with the parameter of

Ravn and Uhlig). Then the cyclical component is obtained by deriving the deviations

of the estimated trend from the actual series.

The actual series, the smoothed series according to HP filter (with the parameter of

Ravn and Uhlig) and the cyclical component for the seven aggregate inequality

indices are presented in the following graphs.

6The parameter using the frequency power rule of Ravn and Uhlig is the number of periods per year

divided by 4, raised to a power, and multiplied by 1600.


13/37

13

-.012

-.008

-.004

.000

.004

.008

.012

.08

.12

.16

.20

.24

60 65 70 75 80 85 90 95 00 05

ATK_05 Trend Cycle

Figure 8. Hodrick-Prescott Filter (lambda=6.25)

-.10

-.05

.00

.05

.10

.15

0.2

0.4

0.6

0.8

1.0

60 65 70 75 80 85 90 95 00 05

ATK_15 Trend Cycle


-.08

-.04

.00

.04

.08

.12

0.2

0.4

0.6

0.8

1.0

60 65 70 75 80 85 90 95 00 05

GE_0_MLD Tre nd Cycle


-.04

-.02

.00

.02

.04

.25

.30

.35

.40

.45

.50

60 65 70 75 80 85 90 95 00 05

GE _1 _THE IL Tre nd C ycle


-.2

.0

.2

.4

.6

0.2

0.4

0.6

0.8

1.0

1.2

60 65 70 75 80 85 90 95 00 05

GE_2_CV Trend Cycle


-.02

-.01

.00

.01

.02

.36

.40

.44

.48

.52

60 65 70 75 80 85 90 95 00 05

GINI Trend Cycle


-.04

-.02

.00

.02

.04

.50

.55

.60

.65

.70

.75

60 65 70 75 80 85 90 95 00 05

RMD Trend Cycle



14/37

14

2.4 International experience

There is an enormous amount of empirical research on income inequality. As a result

several cross-national datasets have been compiled (for a review, see Atkinson and

Brandolini 2001).

One of the most influential projects is the Luxemburg Income Study (LIS). LIS is

covering a variety of countries (mainly European in addition to USA, Australia,

Mexico etc.) and provides micro data for demographic, labor market, expenditure and

income variables. There are six waves7

(around 1980, 1985, 1990, 1995, 2000 and

2004) and wave 7 is under a new template (revised in 2011). The micro data are

derived from various surveys; for example data for Greece are included in Wave IV

(ECHP8

data), Wave V (ECHP data) and Wave VI (EU-SILC9

data). Consequently,

data from national surveys are harmonized and standardized before the calculation ofincome inequality indices.

Another known database is the dataset compiled by Deininger and Squire (1996).

They combined many earlier datasets and evaluated the quality of their observations.

The issue, nevertheless, is that the estimates are based on different income definitions

and reference units; they refer that differences in the definition of the underlying data

might still affect intertemporal and international comparability.

The successor to the Deininger and Squire data set is the World Income Inequality

Database (WIID), created by the United Nations University (UNU-WIDER). Data

from the two previous datasets and from other sources are incorporated. Also the newdata of Deininger and Squire 2004 are included; the update is only published in

WIID2 (current version 2c in 2008) due to agreement between the World Bank and

WIDER to publish one database only. From methodological point of view, definitions

of income and consumption are used, the household is to be the basic statistical unit

(otherwise it is considered problem and is reported) and income or consumption is to

be adjusted to take account of household size using per capita incomes or

consumption.

Babones and Alvarez-Rivadulla (2007) made an effort to unify the data of WIID

compiling the Standardized Income Distribution Database (SIDD). They estimated

adjustment factors for different scopes of coverage, income definitions and reference

7And historical databases for Canada, Germany, Sweden, UK and USA

8 The European Community Household Panel (ECHP) is a survey based on a standardized

questionnaire covering a wide range of topics such as income, health, education etc. The survey was

launched in 1994 and ended at 2001. See also Section 4/ECHP.9The European Union has set up a survey for collecting data on income, poverty, social exclusion and

living conditions. The European Union Survey on Income and Living Conditions (EU SILC) includes

micro data on income on household and personal level that can be used for the estimation of income

distribution. This survey replaced the European Community Household Panel (ECHP). See also Section

3.


15/37

15

units which bring all data to a common standard based on national coverage, gross

income and household per capita inequality.

Another database is the Standardized World Income Inequality Database (SWIID)

compiled by Solt (2009). Solt uses a custom missing data algorithm to standardize the

WIID (ver 2c-2008); data collected by the LIS served as standard. According to theauthor the SWIID provides comparable Gini indices of gross and net income

inequality for 173 countries (SWIID version 3.1 ) for as years as possible from 1960.

Data from SWIID are selected in order to compare the empirical findings from the

usage of tabulated tax data with other countries. It should be noted that the results are

not totally comparable since the data sources and the methodology differs. As it is

implied SWIID utilizes all available data; it summarizes (before the standardizing

algorithm) the reference units (5 categories) and income definitions (4 categories). For

comparison reasons, the choice of the variable is the gross income since the indices

derived from tabulated tax data refers to declared income, that is the tax has not

been paid; nevertheless the amounts does not include social contributions. Therefore,

the substantial existing inconsistencies should be taken into consideration and the

comparison is conducted for indicative reasons.

The following figure presents the Gini coefficient derived from the SWIID and

tabulated tax data for Greece. It is apparent that considerable differences exist. The

time series from SWIID indicate intense variations during the whole period.

Compared to tax data, values are higher until 1993 while values are quite close in the

beginning of 70s in the middle of 80s and in the beginning of the decade of 2000.

Similarities in the trend (increase) are detected in the second half of 80s and 90s,

being more intense in estimates derived from SWIID and tax data correspondingly.

Sources: Authors calculations and SWIID 3.1

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 15 - International Comparison I - Gini coefficient

Greece Greece_tax


16/37

16

The comparison of Gini s estimates (grouped tax data) for Greece is conducted with

two country groups. The first group consists of South European countries such as

Italy, Spain, Portugal and France (although France could be considered part of Central

Europe). The second group includes countries from Central and North Europe

(Germany, Switzerland, Netherlands and Sweden) as well as UK and USA.The results of the comparison of Greece with the first group (Italy, France, Spain,

Portugal) are presented in the next figure. Looking at the whole period, aggregate

income inequality in Greece is usually lower than Portugal, higher than Spain (with

the exception of late 60s and mid-70s), France (with the exception of first half of 90s

and second half of the decade of 2010) while is lower than Italy until 1980 and higher

from mid 90s and onwards. It is noticeable that Gini coefficient is in higher level in

Greece from the mid 1990s with the exception of Portugal and partly France; in

France is higher only in the second half of the last decade.


The outcome of the comparison of Greece with the second group (Germany,

Switzerland, Netherlands, Sweden, UK and USA is presented in the next figure. Until

1980, inequality in Greece is higher than in UK and in the same levels with USA

(though in USA is higher prior to 1970) and lower than other countries with the

exception of certain years (almost equal for Germany in 1972 and 1977, Sweden in

1975 and Netherlands in 1973 and 1977) or periods (lower in Sweden in the late 60s).

In the decade of 1980 inequality in Greece is higher only compared to Netherlands

and partly Germany (only for the first half of the decade) and in the same level with

Switzerland and partly USA and UK (both in the beginning of the decade). Greek

Gini increases more intensely in the beginning of 90s. In the second half of 1990s

aggregate income inequality in Greece is higher than every country. It is exceed onlyby Germany (late 90s) and Netherlands (early 00s). Finally in the second half of the

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

0.65

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 16 - International Comparison II - Gini coefficient

France Italy Spain Portugal Greece_tax


17/37

17

last decade the level are similar to UK and slightly above USA, Sweden and

Switzerland.


Since, as it was noted, the results are not fully comparable it is not necessary to take

into account the absolute levels of the Gini s estimates. The broaderconclusion could

be that after the mid 1990s aggregate income inequality in Greece is in high levelscompared with other countries, while it was a medium case in the previous period.

3. Aggregate measures of income inequality from EU-SILC data

3.1. Estimation of aggregate measures of income inequality from EU-SILC data

The European Union has set up a survey for collecting data on income, poverty, social

exclusion and living conditions. The European Union Survey on Income and Living

Conditions (EU SILC) includes micro data on income on household and personal

level that can be used for the estimation of income distribution. This survey replaced

the European Community Household Panel (ECHP). The EU SILC project was

launched in 2003 for Greece. The data are produced on annual basis and the reference

population is all private households and their current members residing in the territory

of the Member State at the time of data collection. The year of the survey contains

data for the previous year; thus survey for 2010 illustrates information for the year

2009.

EU SILC data contain information for various components of income. Therefore,

several variables that approach the concept of income have been calculated. Thesevariables have been utilized to estimate the distribution of income in the whole

0.25

0.3

0.35

0.4

0.45

0.5

0.55

0.6

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 17 - International Comparison III - Gini coefficient

Germany USA UK Ireland Sweden Switzerland Netherlands Greece_tax


18/37

18

population. Eighteen (18) variables were compiled. The most appropriate - according

to the topic - have been used for income inequality analysis.

These variables describe the concept of income on household level. The size of the

household and the age of its members are important factors, therefore the use of an

equivalence scale is appropriate. In this study the "OECD-modified scale" is

utilized. This scale, first proposed by Haagenars et al. (1994), assigns a value of 1 tothe household head, of 0.5 to each additional adult member and of 0.3 to each child.

The time period of the analysis is from the year 2002 to the year 2009.

The variable used for the estimation of income distribution is the Total net household

income_ no negative PY050N (HY010net_nn). It has been adjusted for the size of

household and the age of the members of household with the OECD-modified scale.

Total net household income_ no negative PY050G (HY010net_nn):

This variable includes net income on household level taking into account, also,components of personal net income. It, therefore, includes net employee cash or near

cash income, company car, net cash benefits or losses from self-employment

(including royalties), unemployment benefits, old-age benefits, survivor' benefits,

sickness benefits, disability benefits, education-related allowances, net income from

rental of a property or land, family/children related allowances, benefit from social

exclusion not elsewhere classified, housing allowances, regular inter-household cash

transfers received, interests, dividends, profit from capital investments in

unincorporated business, income received by people aged under 16.

In this case we do not take into account the negative values in the variable net cash

benefits or losses from self-employment (including royalties).

This variable is slightly different from the corresponding one (Total disposable

household income (HY020)) used by ELSTAT. The rationale was mainly the

conceptual difficulty of incorporating negative income in the compilation.

Nevertheless the latter variable will be used for international comparison purposes.

Basic descriptive statistics, percentiles, income shares (deciles), ratio of income

shares and aggregate income inequality measures have been estimated for the Total

net household income_ no negative PY050N (HY010net_nn). The inequality indices

are Gini coefficient, Atkinson index (parameters 0,5 and 1,5), General Entropy

Indices [parameters 0 (Theil L), 1(Theil T) and 2] and Coefficient of Variation. Theresults are summarized in the following table. It is noted that the years are the

reference periods and not the years the survey has been conducted, i.e 2009 are results

from the survey of 2010.


19/37

19

Table 5

Total net household income_ no negative PY050N (HY010net_nn)

HY010NET_NN_EQ 2002 2003 2004 2005 2006 2007 2008 2009

Observations 6665 6252 5568 5700 5643 6503 7036 7005

Average 9756 10211 10945 11480 12133 12905 13441 13503

St.dev 7500 7471 8127 8719 9744 10165 10827 10176Percentiles-1% 1245 265 1000 1817 1333 1500 0 1440

Percentiles-5% 2400 2732 3000 3251 3433 3889 4014 4260

Percentiles-10% 3344 3760 3945 4153 4336 4861 5043 5357

Percentiles-25% 5300 5699 6000 6188 6526 7125 7619 7676

Percentiles-50% 8012 8566 9000 9415 9847 10600 11013 11147

Percentiles-75% 12203 12902 13587 14180 14770 15822 16373 16420

Percentiles-90% 17300 18040 19600 20513 21684 22590 23125 23400

Percentiles-95% 21793 22310 24667 26012 26871 28000 29174 29005

Percentiles-99% 37667 37727 42830 43975 49068 49424 57400 51997

Shares

Decile_01 0,02 0,02 0,03 0,03 0,03 0,03 0,02 0,03

Decile_02 0,04 0,04 0,04 0,04 0,04 0,04 0,04 0,05Decile_03 0,05 0,06 0,05 0,05 0,05 0,06 0,06 0,06

Decile_04 0,06 0,07 0,07 0,06 0,06 0,07 0,07 0,07

Decile_05 0,08 0,08 0,08 0,08 0,08 0,08 0,08 0,08

Decile_06 0,09 0,09 0,09 0,09 0,09 0,09 0,09 0,09

Decile_07 0,10 0,11 0,10 0,10 0,10 0,10 0,10 0,10

Decile_08 0,13 0,13 0,12 0,12 0,12 0,12 0,12 0,12

Decile_09 0,16 0,16 0,15 0,16 0,15 0,15 0,15 0,15

Decile_10 0,26 0,25 0,26 0,27 0,27 0,26 0,27 0,26

S90/S10 11,11 10,45 10,41 10,05 10,52 9,63 11,56 9,37

S80/S20 6,38 5,98 6,14 6,12 6,21 5,86 6,22 5,57

GINI0,352 0,340 0,348 0,350 0,354 0,343 0,349 0,335

Atkinson 0,5 0,104 0,100 0,102 0,102 0,106 0,101 0,103 0,096

Atkinson 1,5 0,280 0,259 0,269 0,268 0,273 0,267 0,265 0,254

GE(0)=Theil L 0,214 0,193 0,204 0,207 0,211 0,200 0,200 0,190

GE(1)=Theil T 0,216 0,193 0,207 0,213 0,222 0,210 0,213 0,199

GE(2)=CV 0,295 0,268 0,276 0,288 0,322 0,310 0,324 0,284

CV 0,769 0,732 0,742 0,759 0,803 0,788 0,805 0,754


According to the Table 5 for theTotal net household income_ no negative PY050N

(HY010net_nn), the average income illustrates an increasing trend, departing from

9.756 in 2002 and resulting in 13.503 in 2009. The level of 2009 is not lower

compared to the corresponding one of 2008, but the trend of the increase is

considerably less intense since Greece had already entered recession.

The 10% income share yields approximately 26% of the generated income, while the

lower 10% is recipient of approximately 2,5% of income. The indices that indicate the

gap between the income shares of certain portions of population are S80/S20 and

S90/S10, which is simply the ratio between the income share of upper and lower

income classes. There has been a small decrease in both indices; nevertheless the

trend is not stable for the whole period. The decrease is more obvious in the year 2009

especially for S90/S10. This implies that the recession, which is more apparent from

2009, seems to affect more the upper income classes. Nevertheless, since no data areavailable for the rest period of economic crisis this aspect is under scrutiny.


20/37

20

The behavior of the aggregate inequality indices (GINI, Atkinson_0,5, Atkinson_1,5,

General Entropy_0, General Entropy_1, General Entropy_2 and Coefficient of

Variation) is rather stable with miniscule decline. In all cases the absolute values are

slightly changing in both directions (increase or decrease); nevertheless, in all cases a

small decrease is noted from 2008 to 2009. This element, also, implies a minisculedecline in inequality in the beginning of economic recession in Greece. Though, due

to the absence of data for the rest of the period of economic crisis no firm conclusions

can be drawn.

The following figures summarizes the main results for the Total net household

income_ no negative PY050N (HY010net_nn). Figure 18 illustrates the trend of the

average income, Figure 19 shows the trend of lower 10% and 20% and upper 10%

and 20% of income shares, Figure 20 contains the indices of S90/S10 and S80/S20

and finally Figure 21 illustrates the trend of the seven aggregate inequality indices.

Once again it is noted that the reference year of the survey is the previous year (t-1).

Sources: Authors calculations


02000

4000

6000

8000

10000

12000

14000

16000

2002 2003 2004 2005 2006 2007 2008 2009

Figure 18. Total net household income_no negative PY050N (HY010net_nn)

Average Income

Average

0

0.05

0.1

0.15

0.20.25

0.3

2002 2003 2004 2005 2006 2007 2008 2009


Income Shares

Decile_01 Decile_02

Decile_09 Decile_10


21/37

21



Top Income Shares

Chrissis, Livada and Tsakloglou (2011) estimated the top income shares from groupedtax data according to Piketty (2001) approach. The concept of the declared income,

which was the underlying variable, is relatively comparable with the variable of the

Total net household income_ no negative PY050N (HY010net_nn). This variable

contains no negative values and the net components of the income are relative similar

to the declared tax income, since most of these components are to be declared to the

tax authorities. It is reminded that, mainly, net amounts are to be reported in the tax

declarations (i.e. salaries, wages, pensions etc). Table 6 illustrates the results for the

variable Total net household income_ no negative PY050N (HY010net_nn) for 1%,

0,5% and 0,1% top income shares (after applying OECD equivalence scale).

0

2

4

6

8

10

12

14

2002 2003 2004 2005 2006 2007 2008 2009


S90/S10 and S80/S20 indices

S90/S10 S80/S20

0

0.1

0.20.3

0.4

0.5

0.6

0.7

0.8

0.9

2002 2003 2004 2005 2006 2007 2008 2009


Aggregate inequality indices

GINI

Atkinson 0,5

Atkinson 1,5

GE(0)=Theil L

GE(1)=Theil T

GE(2)=CV

CV


22/37

22

Table 6. Top income shares for Total net household income_no negative PY050N(HY010net_nn)

Year 2002 2003 2004 2005 2006 2007 2008 2009

TIS 1% 5,38% 5,01% 5,30% 5,49% 5,99% 5,98% 6,15% 5,80%

TIS 0,5% 3,30% 3,02% 3,11% 3,30% 3,77% 3,71% 3,79% 3,59%

TIS 0,1% 1,06% 1,02% 0,89% 0,96% 1,14% 1,05% 1,20% 1,08%


3.2. International experience

The main variable used in this study for the estimation of income distribution is the

Total net household income_ no negative PY050N (HY010net_nn), which

incorporates the net components of household income without taking into account

negative values for net cash benefits or losses from self-employment (including

royalties). As noted in this section, this variable is slightly different in interpretation

and in compilation procedure from the corresponding one (Total disposable

household income (HY020)) used by ELSTAT, which is described as

Total disposable household income (HY020):This variable includes income on household level taking into account, also,

components of personal income. It, therefore, includes gross employee cash or near

cash income, company car, gross cash benefits or losses from self-employment

(including royalties), unemployment benefits, old-age benefits, survivor' benefits,

sickness benefits, disability benefits, education-related allowances, income from

rental of a property or land, family/children related allowances, social exclusion not

elsewhere classified, housing allowances, regular inter-household cash transfers

received, interests, dividends, profit from capital investments in unincorporated

business and income received by people aged under 16 (HY110G)) minus regulartaxes on wealth, regular inter-household cash transfer paid and tax on income and

social insurance contributions.

The concept of disposable income differs from the concept of net income but this is

the only variable that can be used for international comparison. The ratio S80/S20 and

Gini coefficient will be reviewed. The following table contains the estimations for

these two measures according to Eurostat (data provided by ELSTAT) and according

to authors calculations (for both variables)

Table 7. Aggregate inequality measuresHY020_Eurostat 2002 2003 2004 2005 2006 2007 2008 2009

S80/S20 6,4 5,9 5,8 6,1 6 5,9 5,8 5,6

Gini 34,7 33 33,2 34,3 34,3 33,4 33,1 32,9

HY020 2002 2003 2004 2005 2006 2007 2008 2009

S80/S20 6,5 6,4 6,2 6,2 6,2 5,9 6,2 5,6

GINI 35,1 34,1 34,3 34,6 34,7 33,8 34,3 33,1

HY010NET_NN 2002 2003 2004 2005 2006 2007 2008 2009

S80/S20 6,4 6,0 6,1 6,1 6,2 5,9 6,2 5,6

GINI 35,2 34,0 34,8 35,0 35,4 34,3 34,9 33,5

Sources: Eurostat and authors calculations

Note 1: The year is the reference year (i.e. 2009 means that the survey year is 2010)

There are some differences for total disposable household income in the compilationbut the trend is the same. Difference are also observed between ELSTAT s press


23/37

23

releases (for years 2002 and 200310

) and Eurostat s data. The differences, also, from

the total net household income are small but they exist. Thus, these differences should

be taken into consideration for the international comparison.

The following figures illustrate the ratio S80/S20 and Gini coefficient for totaldisposable household income for Greece and European Union 2711

and Euro Area

1712

. The reason for the sort period for comparison is due to the lack of data for

European averages. It is reminded that years correspond to reference years and not the

years of survey.

Source: Eurostat

10According to press releases for the year 2003 (press release 2004): The S80/S20 is 6,0 (instead of

5,9) and Gini is 33,1 (instead of 33) and for the year 2002 (press release 2003): The S80/S20 is 6,6

(instead of 6,4) and 35,1 (instead of 34,7)11

The European Union (EU27) consists of 27 Member States: Belgium, Bulgaria, the Czech Republic,

Denmark, Germany, Estonia, Ireland, Greece, Spain, France, Italy, Cyprus, Latvia, Lithuania,

Luxembourg, Hungary, Malta, the Netherlands, Austria, Poland, Portugal, Romania, Slovenia, Slovakia,

Finland, Sweden and the United Kingdom plus the European Central Bank and the EU institutions.12

The euro area (EA17) consists of 17 Member States: Belgium, Germany, Estonia, Ireland, Greece,Spain, France, Italy, Cyprus, Luxembourg, Malta, the Netherlands, Austria, Portugal, Slovenia, Slovakia

and Finland plus the European Central Bank.

0

1

2

3

4

5

6

7

2004 2005 2006 2007 2008 2009

Figure 22. S80/S20_Total disposable household income

EU (27 countries) Euro area (17 countries) Greece


24/37

24

Source: Eurostat

The empirical findings indicate that aggregate income inequality in Greece is in

higher level than the average of both European Union and Euro area.

The following figures illustrate analytical results for the year 2009 for the ratio

S80/S20 and Gini coefficient. In both cases Greece yield lower aggregate income

inequality only from Lithuania, Spain, Latvia, Romania and Bulgaria, whereas

inequality is higher in Portugal, Ireland and United Kingdom according to Gini and

lower (Portugal is the same) according to S80/S20 ratio. In any case Greece seems to

suffer from intense aggregate income inequality for the European standards.

Source: Eurostat

26

27

28

29

30

31

32

33

34

35

2004 2005 2006 2007 2008 2009

Figure 23. GINI_Total disposable household income

EU (27 countries) Euro area (17 countries) Greece

0

1

2

3

4

5

6

7

8

EU(27cou

ntries)

Euroarea(17cou

ntries)

Belgium

Bulgaria

CzechRepublic

De

nmark

Ge

rmany

E

stonia

Ireland

Greece

Spain

France

Italy

Cyprus

Latvia

Lithuania

Luxem

bourg

Hungary

Malta

Nethe

rlands

Austria

Poland

Portugal

Ro

mania

Slovenia

Slovakia

F

inland

Sweden

UnitedKingdom

Figure 24. S80/S20_Total disposable household income 2009


25/37

25

Source: Eurostat

4. Results from other data sources

The empirical findings from Household Expenditure Survey (HES) and European

Community Household Panel (ECHP) micro data are presented in this section.

Household Expenditure Survey (HES)

It has been stated that micro data from Household Expenditure Survey (HES) have

been utilized for the estimation of income inequality. According to Mitrakos and

Tsakloglou (2012) available data exist for the HES of 1974, 1981/82, 1987/88,

1993/94, 1998/99, 2004/05 and 2008. The concept of income includes monetary

incomes from all sources, such as wages, self-employment earnings, pensions, rents,

interest payments dividends, cash benefits (net of tax paid). Moreover, the definition

of income includes the non-cash components, namely, imputed rents, other non-cash

incomes (consumption of own farm and non-farm production, in-kind transfers from

other households and fringe benefits). Adjustments were made for the size of the

household; the equivalence scale used was 1,00 for head of household, 0,5 for othermember above 13 years and 0,3 for under 13 years. Moreover, data of each HES are

expressed in constant mid-year prices and then in 1974 constant prices.

It should be noted that the authors compile, also, the distribution of consumption

expenditures and they state that income information from HES is considered less

reliable from ELSTAT. Nevertheless the conclusions do not differ substantially using

the two distributions. Other researchers utilize only consumption data [Sarris and

Zografakis (2000)].

0

5

10

15

20

25

30

35

40

Figure 25. GINI_Total dispasable household income 2009


26/37

26

The empirical results from income distribution are illustrated in the following tables13

Table 8. Top Income Shares from HES micro data

INCOME SHARES 1974 1982 1988 1994 1999 2004 2008

1 2,3 3,2 3,0 3,1 3,0 3,5 3,7

2 4,0 4,9 4,8 4,8 4,7 5,1 5,2

3 5,1 6,0 6,0 5,9 5,9 6,1 6,2

4 6,1 7,0 7,0 7,0 6,8 7,1 7,1

5 7,2 8,0 8,0 8,1 7,9 8,1 8,2

6 8,4 9,1 9,1 9,3 9,0 9,3 9,3

7 9,9 10,4 10,5 10,6 10,4 10,6 10,5

8 12,0 12,2 12,3 12,3 12,1 12,2 12,1

9 15,3 14,8 15,0 14,9 15,0 14,7 14,6

10 29,7 24,3 24,4 24,0 25,1 23,2 23,3

Source: Mitrakos and Tsakloglou (2012)

The aggregate inequality measures are summarized in the following table

Table 9. Aggregate inequality measures from HES micro data

1974 1982 1988 1994 1999 2004 2008

GINI 0,382 0,309 0,314 0,310 0,322 0,292 0,288

VARIANCE OF LOGARITHMS (L) 0,497 0,314 0,339 0,322 0,346 NA NA

THEIL (T) INDEX 0,274 0,170 0,176 0,170 0,187 NA NAMEAN LOGARITHMIC DEVIATION (N) 0,255 0,161 0,170 0,163 0,177 NA NA

ATKINSON INDEX (0,5) 0,123 0,079 0,082 0,079 0,086 NA NA

ATKINSON INDEX (2,0) 0,407 0,274 0,295 0,279 0,300 NA NA

Sources: Mitrakos and Tsakloglou (2012), Mitrakos (2005) (2003) (1999), Mitrakos and Tsakloglou (1998)

Furthermore, Mitrakos and Tsakloglou (2012) estimate the Gini coefficient without

imputed personal income. As expected the coefficient is larger.

Table 10. Gini coefficient from HES micro data

1994 1999 2004 2008

GINI 0,340 0,347 0,325 0,310

Source: Mitrakos and Tsakloglou (2012)

Another interesting aspect is that Mitrakos (2007) has compiled the 1% top incomeshare based on HES data. It should be noted, nevertheless, that the shares in this study

differ slightly from the previous ones presented.Table 11. 1% Top Income Share from HES micro data

UPPER SHARE 1974 1988 1994 2004

1% 2,3 3,0 3,1 3,5

Source: Mitrakos (2007)

European Community Household Panel (ECHP)

The European Community Household Panel (ECHP) is a survey based on a

standardized questionnaire covering a wide range of topics such as income, health,

education etc. The survey was launched in 1994 and ended at 200214

. According to

Eurostat the characteristics of ECHP is the multi-dimensional coverage, the cross-

national comparability and the longitudinal or panel design. The definition of income

refers to total household income. Total household income is taken to be all the net

monetary income received by the household and its members at the time of the

interview (t) during the survey reference year (t-1).This includes income from work

(employment and self-employment); private income (from investments, property and

13The figures coincides with Mitrakos (2007) for the years 1974, 1988, 1994 and 2004, with Mitrakos

(2003) for the years 1982, 1988, 1994 and 1999, with Mitrakos (1999) for the years 1974, 1982, 1988and 1994 and with Tsakloglou and Mitrakos (1998) for 1994.14

Eurostat refers duration of 8 years (1994-2001)


27/37

27

private transfers to the household), pensions and other social transfers directly

received. No account has been taken of indirect social transfers (such as the

reimbursement of medical expenses), receipts in kind and imputed rent for owner-

occupied accommodation. In order to take into account differences in household size

and composition in the comparison of income levels, the amounts given here are perequivalent adult. Thehouseholds total income is divided by its equivalent size,

using the modified OECD equivalence scale. This scale gives a weight of 1.0 to the

first adult, 0.5 to the second and each subsequent person aged 14 and over and 0.3 to

each child aged under 14 in the household. It should be noted that equivalised income

is defined on the household level, so that each person (adult or child) in the same

household has the same equivalised income. The year of the survey contains data for

the previous year; thus survey for 2002 illustrates information for the year 2001.

The empirical findings for the Gini coefficient and for the S80/20 ratio are presented

in the following table:

Table 12. Gini coefficient and ratio S80/20 from ECHP micro data

Year of Survey 1995 1996 1997 1998 1999 2000 2001 2002

GINI 0,35 0,34 0,35 0,35 0,34 0,33 0,33 0,35

S80/20 6,5 6,3 6,6 6,5 6,2 5,8 5,7 6,6

Sources: Eurostat (2002), Eurostat website, ELSTAT various bulletins

Note: Year: Year of survey

5. Comparisons

In the previous sections different data sources and methodological approaches have

been applied for the estimation of income inequality. Moreover, results from other

studies have been presented. The main differences can be categorized as follows

- Data sources: Grouped tax data, Household Expenditure Survey (HES) microdata, European Community Household Panel (ECHP) micro data and

European Union Survey on Income and Living conditions (EU-SILC) micro

have been used

- Methodology: There are certain variations in the methodology applied. Theusage of grouped or micro data dictates the application of different statisticalspecification of the aggregate inequality indices (interpolation techniques

have, also, been used in the case of grouped tax data). Moreover different

compilation procedure was employed in the case of top income shares in tax

data.

- Unit of analysis/ equivalence scale: The unit of analysis is the household in allcases. Nevertheless the equivalence scale is only used when micro data are

available

- Income: The definition of income is not the same; studies using HES includealso items of imputed person income


28/37

28

Despite these differences it is interesting to compare the empirical findings from a

macroeconomic point of view.

The following figure illustrates the results for the estimation of Gini coefficient from

tabulated tax data and micro data from HES, ECHP and EU-SILC.

Sources: Authors calculations, Mitrakos and Tsakloglou (2012) (1998), Mitrakos (2005) (2003), ELSTAT various bulletins,

Eurostat (2000) (2003)

Note 1: Gini_HES_NI: Gini from HES micro data with no imputed personal income items - Mitrakos and Tsakloglou (2012)

Note 2: Gini_HES: Gini from HES micro data - Mitrakos and Tsakloglou (2012) (1998), Mitrakos (2005) (2003)

Note 3: Gini_ECHP: Gini from ECHP micro data - Eurostat (2002), Eurostat website, ELSTAT various bulletins

Note 4: Gini_EU SILC: Gini from EU-SILC micro dataauthors calculations

Note 5: Gini_TAX: Gini from grouped tax dataauthors calculations

The Gini coefficient derived from tabulated tax data (GINI_tax) is in higher level in

all cases. As expected Gini from HES micro data (GINI_HES) yields the smallervalues, since it includes non cash components. Data from HES with no imputed

personal income (GINI_HES_NI) result in higher values of the coefficient. The

coefficient is both lower (1994) and higher (1999) compared with the corresponding

one from ECHP data (GINI_ECHP). Furthermore, Gini is higher (compared to HES

in 2004 and 2008) when is derived from EU-SILC micro data (GINI_EU SILC).

From macroeconomic point of view, apart from the level, the behavior of the

coefficient is important. According to HES data, there is an impressive decrease from

1974 to 1982. For the period 1982-1999 the level of the income inequality does not

alter significantly. On the contrary a decreasing trend exists for the period 1999-

0

0.1

0.2

0.3

0.4

0.5

0.6

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 26. GINI coefficient from various data sources

GINI_HES_NI GINI_HES GINI_ECHP GINI_EU SILC GINI_TAX


29/37

29

200815

. The trend is similar for HES data when imputed personal income is not

included for the period 1994-2008: a small increase is detected for 1994-1999

followed by a small decrease for the remaining period; as already stated the levels of

the coefficient in this case is higher. Micro data from ECHP indicate a relative

constant trend for the period 1994-2001. The coefficient derived from EU-SILC microdata yields a rather constant pattern until 2006 and presents a slight decrease until

2009. The Gini coefficient from tabulated tax data implies an increase of inequality.

The upward trend seems to take place from the early 1990s, being relatively steady in

the previous period.

The pattern is similar for Gini from tax and HES data for the period 1982-1988.

Similarities in the behavior exist for the period 2000-2009 for all cases (with small

variations as described previously).

The following figures illustrate the results for the estimation of the upper shares of

income distribution from tabulated tax data and micro data from HES and EU-SILC.

The 10%, 1% , 0,5% and 0,1% top income shares are presented (only the first two

cases are available for HES data).

Sources: Authors calculations, Mitrakos and Tsakloglou (2012) (1998), Mitrakos (2007) (2003)

Note 1: HES_10%: 10% TIS from HES micro data - Mitrakos and Tsakloglou (2012) (1998), Mitrakos (2003)

15According to Mitrakos and Tsakloglou (2012) this trend is not supported for the period 2004-2008

from the expenditure distribution

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 27. 10% top income shares from various data sources

HES_10% HES2_10% EU-SILC-10% TIS_10%


30/37

30

Note 2: HES2_10%: 10% TIS from HES micro data - Mitrakos (2007)

Note 3: EU SILC_10%: 10% TIS from EU-SILC micro data authors calculations

Note 4: TIS_10%: 10% TIS from grouped tax data authors calculations

The top 10% derived from micro HES data is around 30% in 1974, drops drasticallyin 1982 (24,3%) and then it remains relatively stable for the period 1982-1994

(between 24%-24,3%). A slight increase in 1999 (25,1%) and then a decrease from

2004 onwards (23,2 and 23,3) is detected for the period 1994-2008. In general the

trend for the period 1982-2008 is rather constant. Similar is the trend for HES data

from Mitrakos (2007) with slightly increased values for the years 1974, 1988, 1994

and 2004. Micro data from EU-SILC indicate a relative constant trend (with

successive ups and downs) for the period 2002-2009. The level of 10% top income

share is around 26% with lower value in 2003 (25,3%) and higher value in 2006

(27,4%). The top 10% share derived from tabulated tax data [according to Piketty

(2001) approach] initiates from a value of 21% and ends up around 26,2%. The level

is relatively constant until the late sixties; after this period there is an increase for

some years. From the mid 1970s the share declines and is in the level of 21%-22%

until the end of 1980s. In the beginning of the next decade the income share of the

10% rises exceeding the initial levels. This trend seems to be interrupted in 2002-

2003.

An interesting aspect is that the values between tabulated tax data and micro data

from HES and EU-SILC do not yield such differences as in the case of Gini

coefficient. The level of 10% top share from HES micro data is higher until 1994 and

lower for the remaining period. The corresponding values derived from EU-SILCmicro data are in lower level for 2002-2005 and higher for 2006-2009. Moreover, EU-

SILC values are above HES values both in 2004 and 2008 (years that HES data are

available).

The following figure illustrate the empirical findings for the 1%, 0,5% and 0,1% of

top income shares.


31/37

31

Sources: Authors calculations and Mitrakos (2007)

Note 1: HES_1%: 1% TIS from HES micro data - Mitrakos (2007)

Note 2: EU SILC_1%-0,5%-0,1%: 1% - 0,5% - 0,1% TIS from EU-SILC micro dataauthors calculations

Note 3: TIS_1%-0,5%-0,1%: 1% - 0,5% - 0,1% TIS from grouped tax dataauthors calculations

The 1% top share from HES data is 7,8% in 1974 and drops to 5,5% in 1982. It

remains virtually unchanged for 1984-1988 (5,4%) and it decrease for the period

1988-2004 (4,5%). EU-SILC data indicate a small decrease from 2002 to 2003 and

then a gradual increasing trend which seems to be interrupted in 2008. The top 1%

share from tabulated tax data initiates from a value of 7,5% and ends up around

5,65%. The level is relatively constant until the late sixties; after this period a slow

but steady decline emerges. This trend remains until the beginning of 1980s; during

this decade the top 1% is around 4%. In the beginning of the next decade the income

share of the 1% rises without nevertheless reaching the initial levels. This trend seems

to be interrupted in 2002-2003.Once again, the values from HES compared to tax data are in higher level for the

period 1974-1994. The trend is similar in this period; both empirical findings indicate

a decrease from 1974 to 1982 and then a relative constant pattern for 1982-1994.

Nevertheless, tax data suggests an increase afterwards while HES data indicate a

further decrease. Data from EU-SILC yield a different pattern compared to the tax

data for the period 2002-2009 despite the fact that values are quite similar for 2006-

2007 and 2009.

The 0,5 % and 0,1% top income shares are available only for tax and EU-SILC data.

The pattern differs for 0,5% upper share until 2006 (decrease for tax data and increasefor EU-SILC data); 2007 onwards pattern and values are similar.

0

0.01

0.02

0.03

0.04

0.05

0.06

0.07

0.08

0.09

1957

1958

1959

1960

1961

1962

1963

1964

1965

1966

1967

1968

1969

1970

1971

1972

1973

1974

1975

1976

1977

1978

1979

1980

1981

1982

1983

1984

1985

1986

1987

1988

1989

1990

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

Figure 28. TIS 1%-0,5%-0,1% from various data sources

HES_1% EU-SILC-1% EU-SILC-0,5% EU-SILC-0,1% TIS_1% TIS_0,5% TIS_0,1%


32/37

32

The behavior of 0,1% does not differ. Both pattern and values for the period 2006-

2009 are quite comparable.

6. Conclusions

This section provides empirical evidence for income inequality in Greece. Various

data sources and statistical techniques have been used for the compilation of

aggregate and disaggregate measures of income inequality. Furthermore, empirical

findings from other studies have been presented and compared.

Tabulated tax data for the period 1957-2009 have been utilized for the compilation of

aggregate income inequality measures. Tax data provide detailed information on

nominal family income and its sources, as reported annually in tax declaration forms.

Family income is the sum of income received by the husband and/or wife. This

definition also includes single persons.

Taking into consideration the issues addressed for the data seven indices have been

estimated16: Gini Coefficient (G), Relative Mean Deviation (M), Atkinson Index

() (=0,5), Atkinson Index () (=1,5), General Entropy (0) [GE(0) Theils L

or Mean Log Deviation) (a=0)], General Entropy (1) [GE(1) Theils T) (a=1)] and

General Entropy (2) [GE(2) monotonic transformation of Coefficient of Variation -

CV) (a=2)].

According to the empirical findings, six indices indicate an increase of incomeinequality while one (GE (2)) indicates the opposite (decrease). The mathematical

results are similar to previous studies [Livada (1988), Livada (1991), Livada and

Tsakloglou (1993), Dimelis and Livada (1994)], but the conclusions differ due to the

quite different reference period. The OLS models with correction term AR(1) do not

face significant issues with autocorrelation and heteroscedascicity. All summary

inequality measures, except GE(2), indicate an upward trend for the period 1957-

2009, whereas GE(2) indicate a decline followed by an increase (explaining thus the

quadratic model of description). Nevertheless, the value of GE(2) never reached its

initial level.Our results were compared with data from Standardized World Income Inequality

Database (SWIID) compiled by Solt (2009). It should be noted that the results are not

totally comparable since the data sources and the methodology differs and the

substantial existing inconsistencies should be taken into consideration. The

comparison of Gini s estimates for Greece is conducted with two country groups. The

first group consists of South European countries such as Italy, Spain, Portugal and

France (although France could be considered part of Central Europe). The second

group includes countries from Central and North Europe (Germany, Switzerland,

16Interpolation techniques according to Cowell and Mehta (1982) were applied in order to tackle the

issue of grouped data


33/37

33

Netherlands and Sweden) as well as UK and USA. The broader conclusion could be

that after the mid 1990s aggregate income inequality in Greece is in high levels

compared with other countries, while it was a medium case in the previous period.

Another data source is the European Union Survey on Income and Living Conditions(EU SILC). This survey includes micro data on income on household and personal

level that can be used for the estimation of income distribution. EU SILC data contain

information for various components of income. Therefore, several variables that

approach the concept of income have been calculated. These variables have been

utilized to estimate the distribution of income in the whole population. Eighteen (18)

variables were compiled. The most appropriate - according to the topic - have been

used for income inequality analysis. The variable used for the estimation of income

distribution is the Total net household income_ no negative PY050N

(HY010net_nn). It has been adjusted for the size of household and the age of the

members of household with the OECD-modified scale. This scale, first proposed by

Haagenars et al. (1994), assigns a value of 1 to the household head, of 0.5 to each

additional adult member and of 0.3 to each child. The time period of the analysis is

from the year 2002 to the year 2009.

According to the empirical results, the average income illustrates an increasing trend,

departing from 9.756 in 2002 and resulting in 13.503 in 2009. The level of 2009 is

not lower compared to the corresponding one of 2008, but the trend of the increase is

considerably less intense since Greece had already entered recession.

The 10% income share yields approximately 26% of the generated income, while the

lower 10% is recipient of approximately 2,5% of income. The indices that indicate thegap between the income shares of certain portions of population are S80/S20 and

S90/S10, which is simply the ratio between the income share of upper and lower

income classes. There has been a small decrease in both indices; nevertheless the

trend is not stable for the whole period. The decrease is more obvious in the year 2009

especially for S90/S10. This implies that the recession, which is more apparent from

2009, seems to affect more the upper income classes. Nevertheless, since no data are

available for the rest period of economic crisis this aspect is under scrutiny.

The behavior of the aggregate inequality indices (GINI, Atkinson_0,5, Atkinson_1,5,

General Entropy_0, General Entropy_1, General Entropy_2 and Coefficient of

Variation) is rather stable with miniscule decline. In all cases the absolute values are

slightly changing in both directions (increase or decrease); nevertheless, in all cases a

small decrease is noted from 2008 to 2009. This element, also, implies a miniscule

decline in inequality in the beginning of economic recession in Greece. Though, due

to the absence of data for the rest of the period of economic crisis no firm conclusions

can be drawn.

The variable used for the estimation of income distribution is slightly different in

interpretation and in compilation procedure from the corresponding one (Total

disposable household income (HY020)) used by ELSTAT, since it incorporates the

net components of household income without taking into account negative values for

net cash benefits or losses from self-employment (including royalties). The later is


34/37

34

used in the international comparison for consistency reasons17, even though the impact

of the excluded element is minuscule.

The ratio S80/S20 and Gini coefficient for total disposable household income for

Greece and European Union 2718 and Euro Area 1719 are compared. The reason for the

sort period for comparison is due to the lack of data for European averages. Theempirical findings indicate that aggregate income inequality in Greece is in higher

level than the average of both European Union and Euro area.

Furthermore analytical results for the year 2009 for the ratio S80/S20 and Gini

coefficient are presented. In both cases Greece yield lower aggregate income

inequality only from Lithuania, Spain, Latvia, Romania and Bulgaria, whereas

inequality is higher in Portugal, Ireland and United Kingdom according to Gini and

lower (Portugal is the same) according to S80/S20 ratio. In any case Greece seems to

suffer from intense aggregate income inequality for the European standards.

Empirical findings from studies that utilize Household Expenditure Survey (HES) and

European Community Household Panel (ECHP) micro data are, also, presented. In

both cases income data are used20. Despite the differences (data sources,

methodological differences such as compilation procedure, unit reference/equivalence

scale, definition of income) a comparison was conducted for empirical findings;

specifically for the Gini coefficient and the top income shares.

The Gini coefficient derived from tabulated tax data is in higher level in all cases. As

expected Gini from HES micro data yields the smaller values, since it includes non

cash components. Data from HES with no imputed personal income result in higher

values of the coefficient. The coefficient is both lower (1994) and higher (1999)compared with the corresponding one from ECHP data. Furthermore, Gini is higher

(compared to HES in 2004 and 2008) when is derived from EU-SILC micro data.

Apart from the level, the behavior of the coefficient is important. According to HES

data, there is an impressive decrease from 1974 to 1982. For the period 1982-1999 the

level of the income inequality does not alter significantly. On the contrary a

decreasing trend exists for the period 1999-200821. The trend is similar for HES data

when imputed personal income is not included for the period 1994-2008. Micro data

17 Small differences exist between ELSTAT and Eurostat data (see section 3.4)18

The European Union (EU27) consists of 27 Member States: Belgium, Bulgaria, the Czech Republic,

Denmark, Germany, Estonia, Ireland, Greece, Spain, France, Italy, Cyprus, Latvia, Lithuania,

Luxembourg, Hungary, Malta, the Netherlands, Austria, Poland, Portugal, Romania, Slovenia, Slovakia,

Finland, Sweden and the United Kingdom plus the European Central Bank and the EU institutions.19

The euro area (EA17) consists of 17 Member States: Belgium, Germany, Estonia, Ireland, Greece,

Spain, France, Italy, Cyprus, Luxembourg, Malta, the Netherlands, Austria, Portugal, Slovenia, Slovakia

and Finland plus the European Central Bank.20

Mitrakos and Tsakloglou (2012) compile, also, the distribution of consumption expenditures and

they state that income information from HES is considered less reliable from ELSTAT. Nevertheless

the conclusions do not differ substantially using the two distributions. Other researchers utilize only

consumption data [Sarris and Zografakis (2000)].21

According to Mitrakos and Tsakloglou (2012) this trend is not supported for the period 2004-2008

from the expenditure distribution


35/37

35

from ECHP indicate a relative constant trend for the period 1994-2001. The

coefficient derived from EU-SILC micro data yields a rather constant pattern until

2006 and presents a slight decrease until 2009. The Gini coefficient from tabulated tax

data implies an increase of inequality. The upward trend seems to take place from the

early 1990s, being relatively steady in the previous period.The pattern is similar for Gini from tax and HES data for the period 1982-1988.

Similarities in the behavior exist for the period 2000-2009 for all cases (with small

variations as described previously).

An interesting aspect is that the values of top income shares between tabulated tax

data and micro data from HES and EU-SILC do not yield such differences as in the

case of Gini coefficient. The pattern of tax and HES data is rather similar for the

period 1974-1994, with the values in the latter case being in higher level; the trend,

nevertheless differs in the remaining period. The trend for tax and EU-SILC data is

not the same for the limited years both data exist, but in the latter years (after 2006)

the estimates are quite similar.

To sum up we could say that empirical evidence from tabulated tax data indicates an

increase on aggregate income inequality. This view is not supported by estimates

derived from other data sources (i.e. HES). The level of aggregate inequality differs

from other empirical results. These findings imply that different data sources and/or

methodological approaches could lead to different conclusions for the direction and/or

level of aggregate income inequality. Finally top income shares yield similar trend

(for certain periods) and level (to the possible extend) regardless the data sources.

This view is consistent with Leigh (2007) that top income shares may be a useful

substitute for other measures of inequality.


36/37

36

References

Atkinson, A.B., and Brandolini, A., (2001), Promise and pitfalls in the use of secondarydata-sets: income inequality in OECD countries as a study case, Journal of Economic

Literature 39 (3): 771-799

Babones, S., and Alvarez-Rivadulla, M.J., (2007), Standardized income inequality data foruse in cross-national research, Sociological Inquiry 77 (1): 3-22

Breusch, T.S., Testing for autocorrelation in dynamic linear models, Australian Economicpapers, vol. 17, 1978, pp.334-355

Chow G.C., (1960), Tests for equality between sets of coefficients in two linear regressions,Econometrica, vol. 28, no 3, 1960, pp. 591-605

Chrissis, K., Livada, A. and Tsakloglou, P., (2011), Top Income Shares in Greece: 1957-2009, Technical Report 256, Department of Statistics, AUEB.

Cowell, F.A. and Mehta, F., (1982), The estimation and interpolation of inequality measures,Review of Economic Studies, p. 273-290

Deininger, K. and Squire, L., (1996), A new data set measuring income inequality, WorldBank Economic Review 10 (3): 565-591

Dimeli, S., and Livada, A., (1994), Business cycles characteristics of income inequality:Evidence from Greece, Athens University of Economics and Business, Department ofStatistics Discussion Technical Report No 6

Dimeli, S., and Livada., A., (1997), Business cycles and income inequality in Greece in A.Kintis (edit.) The presence and the future of Greek Economy, Vol B, p. 421-448, Athens

University of Economics and Business, Athens

Durbin, J. and Watson, G. S, (1951), Testing for serial correlation in least-squaresregression, Biometrika, vol. 38, 1951, p.159-171

Godfrey, L. G., Testing against general autoregressive and moving average error modelswhen the regressors include lagged dependent variables, Econometrica, vol. 46, 1978, pp.1293-1302

Hagenaars, A., K. de Vos and M.A. Zaidi (1994), Poverty Statistics in the Late 1980s:Research Based on Micro-data, Office for Official Publications of the European

Communities. Luxembourg.

Hodrick, R.J., and Prescott, E.C., (1980): Post-war US business cycles: an empiricalinvestigation. Manuscript, Carnegie-Mellon University

Jarque, C.M., and Bera, A. K., (1987), A test for normality of observations and regressionresiduals, International Statistical review, vol. 55, 1987, pp 163-172

Leigh, A., (2007), How closely do Top Income Shares track other measures of inequality,The Economic Journal 117

Livada, A., (1988), Aspects of income inequality in Greece, Phd Thesis, University of Essex Livada, A., (1991), Income inequality in Greece: a statistical and econometrical analysis,

Oxford Bulletin of Economics and Statistics 53, p 69-82

Livada, A., and Tsakloglou, P., (1993), Economic inequality in Greece: Time change andstructural analysis, in N. Petralias (edit.), Dimensions of Social Politics today, Institution of

Saki Karagiorga, Athens

Mitrakos, Th., (1999), Aspects of income inequality in Greece, Phd Thesis, Athens UniversityOf Economic And Business Mitrakos, Th., (2005), Estimations of inequality and poverty: statistical techniques and

problems, Greek Statistical Institute, Minutes of 18th Greek Statistical Conference, p. 267 -

274

Mitrakos, Th., (2007), Study of income inequality in Greece with Kernel estimators, GreekStatistical Institute, Minutes of 20th Greek Statistical Conference, p. 267-275

Mitrakos, Th., and Tsakloglou, P., (1998a), Changes in inequality and poverty in Greece after1974, Athens University of Economics and Business, Department of Economic Studies,

Discussion Paper N 98-05

Mitrakos. Th. (2003), Measurement techniques for economic inequality: An application inGreece for the last 20 years, in Th. Skoutzos (ed.) Volume of essays in honour of Professor

Apostolos Lazaridis, p. 249-287, University of Piraeus

Mitrakos. Th. and Tsakloglou, P., (2012) Inequality, poverty and social welfare in SocialPolicy and Social Coherence in Greece during Economic Crisis Conditions, Bank of Greece


37/37

Piketty, T. (2001), Les hauts revenus en France au 20me siecle . Ingalits et redistributions,1901-1998, Editions Grasset

Ravn, M.O. and Uhlig H. (2002), On Adjusting the Hodrick-Prescott Filter for the Frequencyof Observations Review of Economics and Stat

Documents

Technical Report 262- Income Inequality in Greece