Quality Review of the OECD Database on Household Incomes and Poverty and the OECD Earnings Database Part III

Jobs, Wages and Inequality: Work Package 1 -

Quality review of the OECD database on household incomes and poverty and the OECD earnings database Part III

VS/2012/0052 (SI2.625180)

20 December 2012

Social Policy Division

http://www.oecd.org/social/inequality.htm

Directorate for Employment, Labour and Social Affairs

1 TABLE OF CONTENTS

PART III. INCOME DISTRIBUTION DATA – COUNTRY REVIEWS ...... 4 INCOME DISTRIBUTION DATA REVIEW – AUSTRALIA ...... 5 INCOME DISTRIBUTION DATA REVIEW –AUSTRIA...... 12 INCOME DISTRIBUTION DATA REVIEW – BELGIUM ...... 20 INCOME DISTRIBUTION DATA REVIEW – CANADA ...... 27 INCOME DISTRIBUTION DATA REVIEW - CHILE ...... 34 INCOME DISTRIBUTION DATA REVIEW – CZECH REPUBLIC ...... 42 INCOME DISTRIBUTION DATA REVIEW – DENMARK ...... 48 INCOME DISTRIBUTION DATA REVIEW – ESTONIA ...... 59 INCOME DISTRIBUTION DATA REVIEW – FINLAND ...... 65 INCOME DISTRIBUTION DATA REVIEW - FRANCE ...... 71 INCOME DISTRIBUTION DATA REVIEW –GERMANY ...... 77 INCOME DISTRIBUTION DATA REVIEW – GREECE ...... 86 INCOME DISTRIBUTION DATA REVIEW – HUNGARY ...... 93 INCOME DISTRIBUTION DATA REVIEW - IRELAND ...... 99 INCOME DISTRIBUTION DATA REVIEW – ICELAND ...... 108 INCOME DISTRIBUTION DATA REVIEW – ISRAEL ...... 114 INCOME DISTRIBUTION DATA REVIEW – ITALY ...... 120 INCOME DISTRIBUTION DATA REVIEW - JAPAN ...... 131 INCOME DISTRIBUTION DATA REVIEW –KOREA ...... 143 INCOME DISTRIBUTION DATA REVIEW – LUXEMBOURG ...... 149 INCOME DISTRIBUTION DATA REVIEW - MEXICO ...... 155 INCOME DISTRIBUTION DATA REVIEW –NETHERLANDS ...... 163 INCOME DISTRIBUTION DATA REVIEW – NEW ZEALAND ...... 170 INCOME DISTRIBUTION DATA REVIEW – NORWAY ...... 177 INCOME DISTRIBUTION DATA REVIEW – POLAND ...... 184 INCOME DISTRIBUTION DATA REVIEW – PORTUGAL...... 192 INCOME DISTRIBUTION DATA REVIEW – SLOVAK REPUBLIC ...... 202 INCOME DISTRIBUTION DATA REVIEW – SLOVENIA ...... 209

2 INCOME DISTRIBUTION DATA REVIEW – SPAIN ...... 215 INCOME DISTRIBUTION DATA REVIEW – SWEDEN ...... 223 INCOME DISTRIBUTION DATA REVIEW – SWITZERLAND ...... 229 INCOME DISTRIBUTION DATA REVIEW - TURKEY ...... 242 INCOME DISTRIBUTION DATA REVIEW–UNITED KINGDOM ...... 250 INCOME DISTRIBUTION DATA REVIEW – UNITED STATES ...... 256

PART III. INCOME DISTRIBUTION DATA – COUNTRY REVIEWS

The 34 country income distribution data reviews – one per OECD member country – compare the OECD benchmark series for income distribution and poverty reporting with alternative estimates which are available on a national or international basis. The first section describes the data sources which underlie the OECD and alternative series, including a table which lists the survey characteristics of these sources. The second section presents a series of charts which shows trends in different available income inequality and income poverty indicators and discusses those. The third section compares the shares of different income components, in particular public transfers and compares these with alternative survey estimates (if available) as well as the OECD data on social expenditures in the case of transfers. The fourth and fifth sections resume and discuss metadata which could explain differences and inconsistencies in results.

4 INCOME DISTRIBUTION DATA REVIEW – AUSTRALIA

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Australia are provided by the Australian Bureau of Statistics, based on the Survey of Income and Housing (SIH). The SIH was conducted annually from 1994–95 to 1997–98, and then in 1999–2000, 2000–01 and 2002–03. The Household Expenditure Survey (HES) was integrated with the SIH for the first time in 2003–04. The SIH is now conducted every two years and is integrated with HES every six years. SIH was collected in 2005–06 and 2007–08, and the HES was integrated with the SIH for the second time in 2009–10. Currently, the OECD income distribution database contains data for the following (income) years: 1994-95, 1999-00, 2003-04, 2007-08, 2009-10. Data which were available and included for earlier years – 1975-76 and 1984 – have been withdrawn from the OECD database on request from ABS as they were considered not comparable with later series.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

The Australian Bureau of Statistics presents estimates of the income and other characteristics of households, including estimates of the distribution of income across the population, in its publication Household Income and Income Distribution (http://www.ausstats.abs.gov.au/ausstats/subscriber.nsf/0/DBE855896D8CA36DCA2578FB0018533C/$Fi le/65230_2009-10.pdf). These are also based on the SIH. The ABS series is accompanied by a note stating that “Estimates presented from 2007–08 and 2009–10 are not directly comparable with estimates for previous cycles due to the improvements made to measuring income introduced in the 2007–08 cycle. Estimates for 2003–04 and 2005–06 have been recompiled to reflect the new treatments of income, however not all new components introduced in 2007–08 are available for earlier cycles in the 2007–08 cycle.” The same note accompanies the ABS provision of OECD reference series.

The Australian Government’s Department of Families, Housing, Community Services and Indigenous Affairs (FaHCSIA) initiated and funds The Household, Income and Labour Dynamics in Australia (HILDA) Survey. This is a household-based panel study which began in 2001. Responsibility for the design and management of the survey rests with the Melbourne Institute of Applied Economic and Social Research (University of Melbourne) who produce annual reports and statistical reports based on the data (http://www.melbourneinstitute.com/hilda/).

1.2.2 International reporting:

LIS, as OECD, uses SIH as a basis for data reporting (see http://www.lisdatacenter.org/our-data/lis- database/by-country/australia/).

The below table presents the main characteristics of the datasets:

5 Table 1. Characteristics of datasets, Australia

Survey of Income and Household Household, Income and Housing (SIH) Expenditure Statistics Labour Dynamics in (HES) Australia Survey (HILDA) Name of the Australian Bureau of Australian Bureau of Melbourne Institute of responsible agency Statistics (ABS) Statistics (ABS) Applied Economic and Social Research Year (survey and Annually from 1994–95 Every six years since Annually since 2001 income/wage) to 1997–98, then in 2003-04 1999–2000, 2000–01 2002–03, and biannually since 2003-04 Period over which Estimates are calculated Disposable income in the income is assessed as weekly income and year ended 30 June of the annualized (for the year year of the wave, e.g. 2001 ended 30 June) by for wave 1. multiplying by 52.14. Covered population Usual residents of Usual residents of Similar to ABS definition private dwellings in private dwellings in except that boarding school urban and rural areas of urban and rural areas of residents, university Australia. 98% of the Australia. 98% of the students, military personnel people living in people living in residing in private dwellings Australia. Australia. are included. Sample size 18 071 households 9 774 households (2009) 2 153 households (wave 11) (2009) Sample procedure Interview from usual Interview from usual multi-staged cluster residents of residents of stratified sampling private dwellings in private dwellings in urban and rural areas of urban and rural areas of Australia. A stratified, Australia. A stratified, multistage cluster multistage cluster design. design. Response rate 84% 75.00% 71.70% Imputation of yes yes yes missing values Unit for data Individual 15+ Individual 15+ Individual 15+ collection Break in series 2007-08? See above Web source: http://www.abs.gov.au/a http://www.abs.gov.au/a http://www.melbourneinstitu usstats/[email protected]/dossby usstats/[email protected]/mf/653 te.com/hilda/default.html title/F0CDB39ECC0927 0.0/ 11CA256BD00026C3D

5?OpenDocument

6 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The OECD reference series of Gini coefficients shows that income inequality in Australia remained fairly flat between 1995 and 2004. After this it began to rise until 2008 before flattening out again to 2010. The national series (also based on SIH) shows a similar trend. The exception being a small dip between 1995 and 1997 followed by a rise to 2000, rather than the flatter trend of the OECD reference series. However, this extra variation is to be expected given the greater number of data points in the national series. For the years where we have data in both series the difference between the two is around 0.006 or 0.007. However, in 2004 the difference is slightly larger and in 2008, the two series match exactly. As a result of this, the national series shows a greater decrease in income inequality between 2008 and 2010, compared to the flatter OECD series.

The national series from the HILDA survey follows that of the SIH national series very closely for the period 2001 to 2004. However, after 2004 the two series start to diverge, with HILDA showing lower values of Gini coefficients. In particular, the HILDA series is much lower in 2008 at 0.312, compared to 0.336 in both OECD and national series from SIH. Further, the HILDA series drops sharply in 2009. Unfortunately, we do not have a corresponding data point for the SIH series as the survey was not run in 2009

Figure 1. Gini coefficients, Australia

OECD Income Distribution Database (SIH) National series (SIH)

National series (HILDA) LIS 0.40

0.35

0.30

0.25 1990 1995 2000 2005 2010

The OECD reference series shows that the P90/P10 ratio for Australia generally increased between 1995 and 2010, with a slight dip in 2004. The national series based on the SIH shows as very similar pattern, as would be expected, with only some slight extra variation in the series resulting from the additional number of data points in that series. However, despite the similar trend there is a gap of around 0.3 between the two series for the years where we have data for both, with the OECD series consistently higher. As with the two series of Gini coefficients, the gap in 2008 is smaller than in other years giving the

7 appearance of a fall between 2008 and 2010 in the national series compared to the flatter OECD series. The LIS series, also based on SIH, matches the OECD reference series almost exactly.

The series from HILDA follows the national series from SIH very closely. In most of the years for which we have the data for both series the difference between the two is less than 0.1. As with the series of Gini coefficients we see a fairly sharp drop between 2008 and 2009. However, the previous data points do not suggest a divergence of the two series in the same way as in the series of Gini coefficients.

Figure 2. P90/P10 Ratio, Australia

2.1.2 Time series of poverty rates

The OECD reference series of income poverty rates (at 50% of median equivalised income) shows rates increasing steadily between 1995 and 2008. They then flatten out between 2008 and 2010. The series of poverty rates from HILDA is much more jagged since there are a greater number of data points available over the time period. There are only two common data points between the series but they are quite close – 13.2% for the OECD reference series and 12.7% for the HILDA in 2004 and 14.6% and 14.2% in 2008. The LIS series matches the OECD series fairly closely but the most recent data is for 2003.

ABS does not publish a series of poverty rates from SIH at either the 50% or 60% threshold. However, the Australian Council of Social Services (ACOSS) has recently produced a report entitled Poverty in Australia which includes a time series of poverty rates at the 50% threshold based on the SIH data, so in this case “National Series (SIH)” refers to data taken from this report. According to the report the figures for 2003-04 are on different income basis to those from 2005-06 and 2007-08, and the 2009-10 is on yet another income basis. In this series there are three common data points with the OECD series. In 2007-08 the two figures match exactly, in 2003-04 the series published by ACOSS is 1.3 percentage points lower (and lower also than the HILDA series.) The largest difference is in the 2009-10 figures where the OECD series is 1.6% points higher. Thus we have the HILDA series showing a decline in poverty rates between 2008 and 2009, the series from ACOSS showing a decline between 2007-08 and 2009-10 but the OECD series remaining almost flat between 2007-08 and 2009-10.

8 Figure 3. Poverty rates (50% of median income), Australia

OECD Income Distribution Database - disposable income National series (SIH) National series (HILDA) LIS 16.0%

15.0%

14.0%

13.0%

12.0%

11.0%

10.0%

9.0%

8.0% 1990 1995 2000 2005 2010

For interpreting relative income poverty rates, the ABS data provision to OECD specifies: “The social security system in Australia results in an income distribution that is very sensitive to any cut-offs defined as a percentage of median income, particularly around the 50% and 60% level. The distribution has significant peaks as a result of a large number of people receiving either a single rate pension or a couple rate pension. For example, in 2007-08 14.7% of Australians have equivalised disposable household income less than 50% of the median, but this increases to 15.4% when using a 51% cut-off and 16.8% using a 53% cut-off.” Figure 2.2 below illustrates the distribution and its sensitivity to the cut-offs for 2007-08.

Figure 4. Distribution of equivalised disposable household income, SIH 2007-08

2.2 Wages

See Part II of the present Quality Review.

9 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Unfortunately, such information is not available for the other data sources described in table 1.

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average incomeK SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2010 natcur 29,027 8,249 6,295 43,572 5,716 3,376 5941.566 (8,999) 49,607 % av HDI 59% 88% 12% 7% 12% -18%

Figure 5 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period, except for the two latest years.

Figure 5. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Differences between OECD and ABS series:

Both series are based on the same data source, SIH. The differences between the series are a result of two factors. Firstly, a different equivalence scale is used. The ABS use the ‘modified OECD’ equivalence scale where as the OECD terms of reference specify the square root scale. Further the ABS uses a different methodology for the treatment of negative incomes in that it sets all negative disposable household

10 incomes to zero prior to equivalisation. However, the OECD method is to set negative incomes to zero at the component levels.

According to initial ABS assessment, the majority of the difference was said to be due to the treatment of negative incomes. However, as a result of further investigation, ABS specifies that it is the choice of equivalence scale that is responsible for the larger part of the differences.

Differences between OECD and HILDA series:

Like the ABS results on SIH, the results based on HILDA also use the ‘modified OECD’ equivalence scale and set negative household disposable income to zero prior to equivalisation.

5. Summary evaluation

Overall, the OECD reference series and the ABS series based on SIH compare well, as would be expected given that they are based on the same data source. For the most part differences seem to be adequately explained by the two factors described above, namely the choice of equivalence scale and the treatment of negative income. However, it is worth nothing that the national series from SIH indicate a slight decline in poverty and inequality between 2007-08 and 2009-10 whereas the OECD series are flatter over the same period. This is particularly true for poverty rates.

The comparison with the HILDA results is slightly less certain. The OECD reference series of Gini coefficients is somewhat higher than the HILDA series from 2005 onward when the two series start to diverge. On the other hand, the HILDA series of P90/P10 ratios is very close to the ABS series based on SIH, and also follows the OECD reference series quite closely. For poverty rates the OECD reference series and the HILDA series only have two years in common but for both those years the two series are only different by around 0.5 percentage points.

11 INCOME DISTRIBUTION DATA REVIEW –AUSTRIA

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD Data have been using two types of sources for Austria:

 Microcensus for 1983, 1993, 1999

 EU SILC Survey on Income and Living Conditions since 2004.

The two sources differ largely in terms of definitions, reference population and method of calculations and are strictly not comparable, the break in series appears since 2004.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Austrian MicroCensus: The Microcensus survey collected data on income of employed persons at intervals of about 2 years during the period from 1981 to 1999. In addition to income data, the Microcensus also collected information on the number of hours worked and on various socio-economic characteristics of the individuals taking part in this survey. The sampling frame consists of all dwellings of the Austrian Housing Census, which is performed (together with the population census) every 10 years. To ensure representativity of the sample over time every year a certain amount of dwellings from the annual dwelling construction statistics is added. The sample is rotating for every Microcensus survey.

ECHP: The European community Household Panel has been used of national reporting between the mid-1990s and early 2000s.

EU-SILC: The National Statistical Office Statistik Austria reports data from the EU SILC Survey on Income and Living Conditions since 2005.

1.2.2 International reporting:

Eurostat is also computing indicators in income inequalities and poverty for Austria based on EU SILC (formerly on the basis of ECHP).

Austria is also included in the Luxembourg Income Study Database (LIS). LIS is using the Microsensus for 1987 and 1995 data, the European Community Household Panel (ECHP) survey for 1994, 1997 and 2000 data and, for 2004 (the latest available year on LIS), the Survey on Income and Living Conditions / EU-SILC has been used.

Table 1. presents the main characteristics of the different sources:

12 Table 1. Characteristics of datasets used for income reporting, Austria

OECD income distribution Statistics Austria (EU SILC) Austrian Mikrozensus (microcensus) Household budget survey Eurostat (EU SILC) LIS database database (EU SILC)

Name Survey on Income and Living Survey on Income and Living Austrian Mikrozensus (microcensus) Konsumerhebung Survey on Income and Living II. Austrian Microcensus na Conditions see OECD term of Conditions Conditions 1987H - 1987P reference IV. European Household Panel / AT ECHP 1994 1994H - 1994P Austrian Microcensus na 1995H - 1995P European Household Panel / AT ECHP 1997 1997H - 1997P V. European Household Panel / AT ECHP 2000 2000H - 2000P VI. Survey on Income and Living Conditions EU-SILC 2005 survey 2004 2004H - 2004P

Name of the responsible Statistics Austria Unit Microcensus Eurostat agency (Directorate Population Statistics, Statistics Austria)

Year (survey and Since 2004 Before 2004 1999/2000 and 2004/05 Every year from 2003 onwards. income/wage) Period over which Annual income in the previous income is assessed year, also in case of transfers from public sources 1983, 1993, 1999 Covered population All private (non-group , non- The sampling frame included the total institutional) households in population of household heads. Excluded Austria. from the sample are: · about 10.000 homeless people, · 70.000 people in hospital, care or nursing homes, · 15.000 people in hostels (students, nurses, etc.) · 2.000 children in children’s homes · 3.000 prisoners · 35.000 people in other "institutional" households (Hotels, bed and breakfast accommodations, cloisters, special quarters for immigrant workers; · and self-employed: about 480.000 self- employed (incl. family workers), who are included in the survey, but are not asked about their income

13 Table 1. Characteristics of datasets used for income reporting, Austria (cont) OECD income distribution Statistics Austria (EU SILC) Austrian Mikrozensus (microcensus) Household budget survey Eurostat (EU SILC) LIS database database (EU SILC)

Sample size 1% of total population, Minimum sample size: 4 500 households for cross- sectionnal/3 250 for longitudinal; 8 750 individuals for 6 250 for longitudinal.

Sample procedure Stratification Two-stage design with first stage unit stratification. The first stage is formed by census sections and the second stage by main family dwellings. Rotating sample: a quarter of the sample is renewed each cycle. Sampling frame: Central Population Register (Zentrale Melderegister - ZMR) as of 31.12.2004. Final sample size: The sample includes 16,000 selected dwellings distributed in 2,000 census sections.

Response rate one third non-response for income questions

Imputation of missing Missing values because of item values non-response as well as partial unit non-response are fully imputed.

Unit for data collection The unit of observation of the The unit of observation is the household Individuals and households. survey is the household. and it is possible to detect the relations between the household members. Break in series Since 2000

Web source: See EU-SILC National and EU http://www.statistik.at/web_en/st See EU-SILC National and EU comparative Quality assessment atistics/social_statistics/househo comparative Quality assessment via ld_income/index.html via http://circa.europa.eu/Public/irc/d http://circa.europa.eu/Public/irc/d sis/eusilc/library?l=/quality_asse sis/eusilc/library?l=/quality_asse ssment&vm=detailed&sb=Title ssment&vm=detailed&sb=Title

14 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The OECD considers the time series pre- and post EU SILC for Austria not comparable. In the different OECD publications, the trends are shown from 1993 through 1999, and from 2004 through 2009. 2004 is considered as “break” year. This Data Review is adopting the same approach.

Eurostat and EU-SILC series report the survey year and not the income year. For comparison purposes, the data had been reported one year backward.

Figure 1.1 Trends in Gini coefficient (disposable income)

Source: Austrian Microcensus, Eurostat, EU Survey on Income and Living Conditions, (EU SILC), European Community Household Panel (ECHP), LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

Up to 2004, LIS and OECD series differ in terms of level and trends. LIS refers to ECHP figures whereas OECD refers to Microcensus which are strictly not comparable. Data from the Microcensus underreport income at the bottom and upper end of the distribution income dispersion and therefore this leads to a lower Gini coefficient. The Microcensus is not representative as far as income data are concerned and does not provide a comprehensive picture of household earnings since income for self- employed and family helpers are not included (unless they are partners of dependent employees).

As of 2004, the various series on Gini coefficients for disposal income in Austria are quite similar and are showing similar trends (this is not surprising as all series are based on EU SILC).

Eurostat reports slightly higher estimates for Gini coefficient and the S80/S20 ratio compared to OECD since 2004. One reason could be that OECD uses the square of the household size for the equivalence whereas Eurostat uses the “OECD modified equivalence scale”.

15 Figure 1.2 Trends in S80/S20

Source: Austrian Microcensus, Eurostat, EU Survey on Income and Living Conditions, (EU SILC), European Community Household Panel (ECHP).

OECD, Eurostat and Statistics Austria data on ratio the ratio of S80/S20 are close.

2.1.2 Time series of poverty rates

Figure 2.1 Trends in poverty rates, after taxes and transfers

16 For the period between 1993 and 1999, OECD series show a different (increasing) trend, for both poverty indicators than the LIS data and Eurostat data (based on ECHP) which suggest stability or slight decrease. An explanation is offered by Biffl (2003): “The household survey of 1993 is not adequately capturing the change in the structure of population between 1989 and 1993. This period is characterised by unprecedented numbers of net-inflows of migrants. The migrants tended to fill the ranks of inhabitants at the bottom end of the income scale. A new sample was drawn in 1994, taking account of the changed structure of the population. By 1999, the migrants have been more or less fully integrated, many of them have become naturalised. Both aspects, the difference in the cyclical position and the structural adjustment of the sample survey may account for some of the rise in income inequality and poverty between 1993 and 1999”

The OECD reference series and Statistics Austria show similar trends since 2004 based on the threshold below 60% of the median income despite the fact that the data reported by Statistics Austria are slightly lower.

The picture is similar with regard to child poverty estimates.

Figure 2.2 Trends in child poverty rates, after taxes and transfers

2.2 Wages

See Part II of the present Quality Review.

17 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI)

OECD reference series 2009 natcur 15 265 6 261 25 21 551 1 113 3 653 9 033 -9 624 25 726 % av HDI 59% 84% 4% 14% 35% -37%

According to Statistics Austria, the median disposable household income for 2009 was EUR 29 849, and the equivalised household income for 2009 was EUR 19 886.

Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Despite a break in series for the OECD reference series, both series show similar trends throughout the period, except for the latest year. It can also be seen that the new data source for Austria (EU-SILC) suggests much higher transfer shares than the old series (Mikrozensus)

Figure 3. Trends in shares of public social transfers

18 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

The Microcensus is not representative as far as income data are concerned and does not provide a comprehensive picture of household earnings since self-employed and family helpers are not included.

The income question is answered by some 70% of the respondents, but only half of the persons surveyed with the highest income are willing to reveal their income, imparting a downward bias to wage differentials. Moreover, among those who do answer the income question, respondents tend to adjust downwards high incomes and to adjust upwards low incomes. They show problems at the bottom and upper end of the distribution. This, too, has the effect that the extent of income dispersion is underreported in the Mikrocensus.

There is two additional issues: first, self-employment income is not reported (if the self-employed person is not living with someone having another source of income). And certain income components, such as bonuses and compensation for overtime are underreported or not reported at all. Again, this results in a downward bias in income dispersion in income data collected by the Mikrozensus (Kronsteiner and Wolf 1994).

Methodological differences between the OECD reference series based from EU SILC and Statistics Austria results based on the same source might be due the use of different equivalence scale.

5. Summary evaluation

As explained above, OECD reference series refer to Austrian Microcensus for 1983 through 1999 and EU SILC Survey on Income and Living Conditions since 2004. The two sources differ largely in term of reference population, definitions and method of calculations and are strictly not comparable, the break in series appears since 2004.

Bibliography

Biffl, G. (2003), “Distribution of Household Income in Austria”, mimeo, WIFO

Development of the Distribution of Household Income in Austria, Gudrun Biffl, WIFO Working Papers, No. 293, May 2007

Einkommen_armut_und_lebensbedingungen, Ergebnisse aus EU- SILC 2007, Herausge geben von Statistik Austria.

Einkommen_armut_und_lebensbedingungen, Ergebnisse aus EU- SILC 2006, Herausge geben von Statistik Austria.

Empirica (2009) 36:389–406, DOI 10.1007/s10663-008-9099-7, ORIGINAL PAPER, How large are wage differentials in Austria? Wolfgang Pollan

Statistik Austria: Intermediate Quality Report Relating to the EU-SILC 2010 Operation, Austria, Vienna, 22nd December 2011.

Statistik Austria : eu-indikatoren_zu_armut_und_sozialer_eingliederung_2011

19 INCOME DISTRIBUTION DATA REVIEW – BELGIUM

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Belgium are computed on the basis of three different data series, depending on the period:

 For the years 1983 and 1995, tax records mainly from IPCAL data, i.e. all tax units filing tax declarations (ca. 86% of total population), provided by the Ministry of Finance

 For the years 1995 and 2000: the European Community Household Panel (ECHP), provided by the University of Antwxerp

 For the years 2005 and annually since 2008: EU-SILC 2005/2009

Due to the change in source in 1995, data for 1983 have been interpolated (“spliced”) and the OECD Datacube provides series for both the original and the adjusted values. However, there is a strict break in series due to the change in source to EU-SILC in 2004, with no overlapping year with two data sources available. Data prior to 2004 are strictly not comparable with data from the 2004 onwards. Note that this break is shown in the Figures below between 2000 and 2004.

1.2. National reporting and reporting in other international agencies:

Income distribution and poverty indicators for Belgium are also available from

 The Eurostat’s database on income and living conditions (based on ECHP and EU-SILC annual surveys since 2000 (survey year 2002 missing), see http://epp.eurostat.ec.europa.eu/portal/page/portal/income_social_inclusion_living_conditions/dat a/database .

 The Luxembourg Income Study Database (LIS), using surveys from the University of Antwerp and University of Liege for 1985, 1988, 1992, 1995, 1997 and 2000, the PSBH (Panel Study on Belgian Households) and SEP (Socio-Economic Panel).

 STATBEL using annual data from the ‘Direction générale Statistique et Information économique de la SPF Economie’ from 1990 to 2008.

The below table presents the main characteristics of those four datasets:

20 Table 2. Characteristics of datasets used for income reporting, Belgium

OECD reference National survey (Income) LIS database series since 2005 (EU-SILC) Name EU-SILC 2005/2009 Direction générale Statistique et Panel Study on Belgian Households Information économique de la SPF (PSBH) Economie , Revenus fiscaux

Name of the Eurostat STATBEL Universiteit Antwerpen (UA) / responsible agency Université de Liège (ULg)

Year (survey and EU-SILC 2005-2009 1990-2008 annually 1992, 1996, 1997 and 2000 surveys income/wage) surveys representing representing income for 1991, 1995, income for 2004-2008. 1996 and 1999 ECHP 1996 and 2001 representing income for 1995 and 2000. Period over which Annual income for the Annual income for the all year N-2 Annual income for the all year N-1 income is assessed all year N-1 (need verification)

Covered population All households in All private households in the whole Belgium, except those national territory (incl. collective in collective households households and but excl. institutional ones) institutions Sample size 6132 households (in 3,067 completed household interviews SILC 2010) out of 3,672 in initial sample + 7,850 individuals in completed households, of which 7,367 interviewed (5,610 16+)

Sample procedure stratified survey, cross-sectional survey broken down by rotating since 2005 8 waves during one year - from a random uniform sample on 1999 census Response rate About 66% Around 60% of households contacted finilised all 3 questionnaires and individual diaries Imputation of missing No missing values, None values negative values treated as suggested in the terms of references Unit for data collection Household Household Household and individual

Break in series Data prior to 2004 No cannot be compared with data from 2004 onwards.

Web source: http://stats.oecd.org/In http://statbel.fgov.be/fr/statistiques/ http://www.lisdatacenter.org/wp- dex.aspx?QueryId=26 chiffres/travailvie/fisc/inegalite_de_ content/uploads/our-lis-

068 revenu/ documentation-by-be00-survey.pdf

21 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among total population has slightly increased overall in Belgium between 1983 and 2000, and slightly decreased between 2004 and 2009, the two trends cancelling each other out, in contrast to most other OECD countries which have seen a more steady increase throughout.

For the period up to 2000, also the LIS and the STABEL data show a steady increase, although on significantly lower levels. For the period since 2004, there are no LIS data series. The Eurostat data series which are based on EU-SILC show generally similar levels, although more fluctuation. Also, to the difference of the OECD, Eurostat considers the data between 2000 (ECHP) and 2004 (EU-SILC) comparable. In stark contrast to both OECD and Eurostat series, the STATBEL series shows a steady increase in levels of inequality since 1995.

Figure 6.1 Trends in Gini coefficient (disposable income)

OECD reference series (tax files & ECHP until 2000, EU-SILC since 2004) Eurostat (ECHP and EU-SILC) LIS (PSBH) STATBEL, Revenus fiscaux 0.35

0.30

0.25 Gini (at disposable income) (at disposable Gini

0.20 1980 1985 1990 1995 2000 2005 2010

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference data suggest a slight increase before 2000 and a slight decrease after 2004. This is similar to the Eurostat series, though the latter show much more fluctuation and suggest a double downward leaning “w” trend.

22 Figure 1.7 S80/S20

OECD reference series (tax files & ECHP until 2000, EU-SILC since 2004) Eurostat (ECHP and EU-SILC) 5.0

4.5

4.0 S80/S20

3.5

3.0 1990 1995 2000 2005 2010

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Belgian population living with less than 50% of the median equivalised income (11.550 Euros per year in 2008) oscillated between 9 and 10% of the population in the late 2000s. The EU-SILC series from Eurostat, although fluctuating, shows a similar somewhat lower level. Also, for the period between 1995 and 2000, all series (OECD, Eurostat and LIS) suggest a downward trend. No STATBEL data are available on this indicator.

Figure 2.1 Trends in poverty rates (at 50% median income)

OECD reference series (tax files & ECHP until 2000, EU-SILC since 2004) Eurostat (ECHP and EU-SILC) LIS (PSBH) 16%

14%

12%

10%

Poverty rate (at 50% median income) median (at 50% rate Poverty 2%

0% 1980 1985 1990 1995 2000 2005 2010

The STATBEL published series measures poverty rates at 60% of median income, as opposed to 50%. Using 60% of median income for unit of measure, the STATBEL and EU-SILC series show similar

23 trends with both showing a 14.6% poverty rate in 2009, while the OECD reference survey shows a higher level of poverty rate at 16.3%.

Figure 2.2 Trends in poverty rates (at 60% median income)

OECD reference series (tax files & ECHP until 2000, EU-SILC since 2004) STATBEL, Revenus fiscaux Eurostat (ECHP and EU-SILC) LIS (PSBH) 20%

18%

16%

14%

12% Poverty Rate (at 60% median income) median 60% (at Rate Poverty 10% 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

As for child poverty, the OECD reference series is from 1995 and shows a trend that is generally higher than the LIS series (before 2000) but also the Eurostat series (from 2004).

Figure 2.3 Trends in Child poverty rates (at 50% median income)

OECD reference series (tax files & ECHP until 2000, EU-SILC since 2004) Eurostat (ECHP and EU-SILC) LIS (PSBH) 13% 12% 11% 10% 9%

8% income) 7% 6%

5% Child Poverty rate (at 50% median median 50% (at rate Poverty Child 4% 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review

24 3. Consistency of income components with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers, taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. The average equivalised disposable income of the OECD reference survey and the EU- SILC series are identical since the OECD extracts its data from the EU-SILC database.

Table 3. Shares of income components in total disposable income, OECD reference series

Self Disposable # Year Unit Wages Capital Employment Transfers Taxes income (HDI) OECD reference suvery 2008 natcur 20,426 1,150 2,233 6,610 -7,503 23,100 % av HDI 88% 5% 10% 29% -32%

Figure 3. Trends in shares of public social transfers

10 % transfers in disposable income (OECD reference series) 5 % cash social spending in Net National Income (OECD SOCX) 0 1990 1995 2000 2005 2010

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

The OECD reference series, as well as the LIS series, use the square root of household size, whereas the EU-SILC series and STABEL series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

The OECD reference series show broadly similar trends than the Eurostat series. OECD reference series based from EU-SILC since 2004 show less variation than the EU-SILC series publish in Eurostat.

25 This may be due to the different equivalence scales used. Treatment of missing and negative income values should not play a role as these seem treated the same way. By contrast, the STATBEL series based on revenus fiscaux show different trends when it comes to income inequality compared to both other series – notably a steady upwards trend, including in the second half of the 2000s. On the other hand, reported poverty rates are broadly similar.

26 INCOME DISTRIBUTION DATA REVIEW – CANADA

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD Income Distribution Data for Canada are computed by Statistics Canada and based on the Survey of Labour and Income Dynamics (SLID). SLID is an annual cross sectional and longitudinal survey conducted since 1996. Data from 1976 to 1996 are from the Survey of Consumer Finances (SCF), an annual cross sectional survey. In 2009, Statistics Canada has revised the SCF data for the OECD in order to match SCF and SLID. Canada provides the full questionnaire response on an annual basis starting with the year 1976.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Income distribution and poverty indicators for Canada are also available from the Census which is conducted every five years.

An additional source is the administrative source from The Canada Customs and Revenue Agency (CCRA), T1FF which is a census of all individuals who file taxes, or whose social insurance number (SIN) appears on another family members’ tax file.

As Census does not include information on taxes, a Census after-tax (Census-AT) dataset has been created by Statistics Canada imputing taxes for families using a reduced form approach with estimation of regressions of taxes paid on family characteristics using tax data and then using the estimated coefficients from those regressions to predict taxes paid in the Censuses (see http://publications.gc.ca/Collection/Statcan/11F0019MIE/11F0019MIE2006274.pdf).

1.2.2 International reporting:

LIS uses, such as OECD, SLID as a basis for data reporting (see http://www.lisdatacenter.org/our- data/lis-database/by-country/canada-2/).

OECD earnings indicators for Canada refer to gross weekly earnings distribution for full-time workers from Labour Force Survey, Statistics Canada, V0902_06 (see Annex III).

Table 1 presents the main characteristics of the different sources:

27 Table 1. Characteristics of datasets used for income reporting, Canada

OECD income National survey SCF Survey of Census Administrative source: The Canada National survey Survey of Labour and Income distribution Consumer Finances (SCF) pre (1980,1985,1990,1995,2000,2006, Customs and Revenue Agency Dynamics (SLID) 1996-2011 (Income) database 1996 (Income) 2011) (CCRA) T1FF (the T1 Family File)

Name see SLID Survey of Consumer Finances Survey of Labour and Income Dynamics (SLID) Census (SCF) Name of the responsible Statistics Canada Statistics Canada Statistics Canada agency Year (survey and 1980 available up to 1996 1996-2011 Every five years income/wage) Period over which income is Each year, a panel is interviewed in January (mainly to Shortly after the tax season (sometime assessed collect labour information) and in May (to collect income in May in recent years). information). Covered population The SCF is an annual The survey covers all individuals in Canada, excluding The Census is a survey aimed at T1FF is a census of all individuals who crosssectional survey that targets residents of the Yukon, the Northwest Territories and collecting information on the entire file taxes, or whose social insurance all households in Canada, with the Nunavut, residents of institutions and persons living on population, and is conducted every five number (SIN) appears on another exception of those living in the Indian reserves. Overall, these exclusions amount to years shortly after the tax season family members’ tax file.Population territories, institutions, or on native less than 3 percent of the population. Due to the survey (sometime in May in recent years). coverage rate has been over 95% reserves. The exceptions account coverage error, SLID covered 84 percent of its target since 1992 in T1FF. for less than 3% of the Canadian population. population. Sample size Small: The sample of roughly Small: The sample contains 64,783 persons, 27,843 Large 20% of the population Large 35,000 households is selected as a economic families and 26,745 households. supplement to the April Labour Force Survey (LFS). Sample procedure Cross-sectional estimates, it is also designed for longitudinal analysis. The samples for SLID are selected from the monthly Labour Force Survey (LFS) and thus share the latter’s sample design (stratified, multi-stage design that uses probability sampling).Panels are interviewed for up to 6 years, with new (and overlapping) panels introduced every 3 years. For the income interview, respondents have the option of allowing Statistics Canada to link to their T1 tax files (if possible) in order to collect their income information, thus eliminating the need for an income questionnaire. More than 80% of respondents provide Statistics Canada with the permission to attempt this match, and the income of about 70% of all respondents is obtained from the tax files in this way. Response rate Response rates generally over Slightly higher (80% to 85%) around 80% Imputation of missing values Unit for data collection Break in series New survey after 2011.Survey of Longitudinal and The T1FF has been particularly well- International Study of Adults (LISA) suited for estimates of income at the lower end of the distribution since 1992 given the large number of incentives for lower income families to file taxes.2 On the other hand, the creation of those incentives implies that tax data are unlikely to be based on a consistent sample of observations before versus after 1992. Indeed, given changes from the Child Tax Credit to the Child Tax Benefit in 1993, the 1992 data may also not be comparable to subsequent years.3 This, plus the fact that transfer income was not reported at all before 1989 and not reported consistently before 1992 implies that tax data before 1992 are not comparable. 1993 should be considered as a starting point in establishing trends.

Definition of reference person Families are derived with respect In SLID, it is with respect to the “major income to the “head” of the family, which recipients" gives priority to the husband Definition of household A household is defined as being composed of a person or group of persons who co reside in,or occupy the same dwelling and do not have a usual place of residence else where in Canada or abroad.Household can refer to the economic family defined as a group of two or more persons who live in the same dwelling and are related to each other by blood,marriage,a common law union or adoption.Household can also contain two or more economic families composed of individuals who are unrelated by blood or law but live together in the same dwellin,such as roommates or alodger (non- familyhousehold). An ndividual living alone is considered as one person household. The household members who are temporarily absent (e.g. temporary residents elsewhere) are considered part of their usual household.

Web source: http://www23.statcan.gc.ca/imdb/p2SV http://www.statcan.gc.ca/pub/75f0011x/75f0011x2012001- http://www5.statcan.gc.ca/subject- http://publications.gc.ca/Collection/Statcan/ .pl?Function=getSurvey&SDDS=3502&l eng.htm sujet/result- 11F0019MIE/11F0019MIE2004219.pdf ang=en&db=imdb&adm=8&dis=2 resultat.action?pid=3868&id=3874&lang=en g&type=CENSUSTBL&pageNum=1&more= 0

28 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Income indicators and Gini coefficients based on surveys are lower than those recorded under census or tax data. As response rates are lower in survey data, and non response is more likely in lower-income families, this results to a relative undercoverage at the very bottom of the income distribution and lower- income families are under-represented. The weights in the survey data are adjusted to make the sample more representative of various province-age-sex groups, as well as household and family sizes. However, no specific adjustments are made to correct the non-representativeness of the sample by income level that may be introduced by non-response.

The OECD reference series for Gini coefficients of disposable and market income are quasi identical to the national series based on SLID and very similar to the LIS series which are equally based on SLID (Figures 1.1 and 1.2)

Figure 1.1 Trends in Gini coefficient (disposable income)

OECD reference series (SLID) National Series (SLID) Census-AT After-income and payroll tax income Administrative tax data T1FF LIS (SLID) 0.40

0.35

0.30 Gini (at (at Gini disposableincome)

0.25 1990 1995 2000 2005 2010

Source: Census-AT After-income and payroll tax income: Analytical Studies Branch Research Paper Series: Revisiting Recent Trends in Canadian After-Tax Income by Marc Frenette, David Green and Kevin Milligan Family and Labour Studies Division; Administrative tax data T1FF: Research Paper, Inequality Using Census Data: Analytical Studies Branch research paper series, Rising Income Inequality in the 1990s: An Exploration of Three Data Sources By Marc Frenette, David Green and Garnett Picot. Business and Labour Market Analysis Division; National Series (SLID): Survey of labour and Income Dynamics; LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

29 Figure 1.2 Trends in Gini coefficient (market income)

Source: Census-AT market income: Analytical Studies Branch Research Paper Series: Revisiting Recent Trends in Canadian After- Tax Income by Marc Frenette, David Green and Kevin Milligan Family and Labour Studies Division; National Series (SLID): Survey of labour and Income Dynamics; LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

2.1.2 Time series of poverty rates

Figure 2.1 Trends in poverty rates

Source: Survey of labour and Income Dynamics (SLID); LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

30 Figure 2.2 Trends in Child poverty rates

Source: Survey of labour and Income Dynamics (SLID); LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

Despite the fact that data are both from SLID, poverty rates based on the OECD income distribution database and LIS differ slightly. LIS series are somewhat higher, in particular during the period 1997 through 2004. This is probably due to the fact that Statistics Canada has revised the data for the OECD in order to match SCF and SLID.

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average incomeK SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2010 natcur 25 558 6 761 3 508 35 827 6 185 3 374 6 146 -10 085 41 447 % av HDI 61.7% 86% 15% 8% 15% -24% Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions,

31 incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period, except for the latest year.

Figure 3 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Differences between survey and tax data:

The differences lie more in coverage than in differential reporting between survey and tax data and whether survey data (both SCF and SLID) cover a different population than tax data. Lower-income families are under-represented in survey data. The weights in the survey data are adjusted to make the sample more representative of various province-age-sex groups, as well as household and family sizes. However, no specific adjustments are made to correct the non-representativeness of the sample by income level that may be introduced by non-response.

Tax data is less likely to suffer from income-related response bias since the coverage rate is above 95% (since 1992). And finally, given the larger sample sizes in tax data, inequality estimates are less susceptible to high levels of sampling variability.

In addition there is a break in series in the tax data as data improved in 1993 with the introduction of the Child Tax Benefit and a sudden increase in lower-income filers. Survey data also have a break in the series in the mid to late 1990s in the transition from SCF to SLID.

Differences between survey and Census:

Census has a greater tendancy to accurately represent families in the bottom of the income distribution as this is mandatory and the coverage is much larger than in surveys. However census data do not collect information on taxes.

The response rate in survey data is generally around 80% to 85%, and no adjustments are made to make the samples representative of the Canadian population in various income groups.

32 5. Summary evaluation

When comparing the pre-tax income distribution from survey and tax data with that of Census data, the bottom end of the income distribution in Census data more closely resembles the tax data than the survey data, both in terms of levels and trends.

The Census collects income data in much the same way as the SCF did, yet the coverage rate is much higher (as in the tax data). However as survey data are annual data and include complete information on income composition including taxes, they were chosen as the most appropriate for analysis in trends in household income distribution and poverty in Canada.

Bibliography:

Analytical Studies Branch Research Paper Series: Revisiting Recent Trends in Canadian After-Tax Income by Marc Frenette, David Green and Kevin Milligan Family and Labour Studies Division.

Research Paper, Inequality Using Census Data: Analytical Studies Branch research paper series, Rising Income Inequality in the 1990s: An Exploration of Three Data Sources By Marc Frenette, David Green and Garnett Picot. Business and Labour Market Analysis Division

33 INCOME DISTRIBUTION DATA REVIEW - CHILE

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Chile are provided by the Ministerio de Desarollo Social (Ministry of Social development), from the National Socio-economic Survey (Caracterizacion Socioeconomica Nacional - CASEN), a household survey conducted at national, regional, urban and rural levels, and developed by the Ministerio de Planificación Nacional y Política Económica (MIDEPLAN). It has been carried out for the following years: 1985, 1987, 1990, 1992, 1994, 1996, 1998, 2000, 2003, 2006, 2009 and 2011. Data provided to the OECD refer to the years 1996, 2006, 2009 and 2011.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Alternative data sources to CASEN are:

 The Encuesta Suplementaria de Ingresos (ESI), conducted by the INE

 The Encuesta de Proteccion Social (EPS), conducted by the Ministry of Social Protection

 The Encuesta Financiera de Hogares (EFH) from the Chilean Central Bank

1.2.2 International reporting:

Chilean data on Income and poverty are also available from the The World Bank PovCalNet database (http://data.worldbank.org/country/chile). The World Bank database, as the OECD, uses the CASEN survey for data reporting.

Table 1 below presents the main characteristics of these different sources:

34 Table 4. Characteristics of dataset, Chile

National Socioeconomic Encuesta Suplementaria de Ingresos Encuesta de Encuesta Characterisation Survey Proteccion Financiera de (Encuesta CASEN) Social Hogares Name The National Socio- Encuesta Suplementaria de Ingresos Encuesta de Encuesta economic Survey (Encuesta Proteccion Financiera de de caracterización Social Hogares Socioeconómica Nacional - CASEN) Name of the Data are provided by the National statistical Office (INE) Ministerio de Banco Central responsible Ministerio de Desarollo Protección de Chile agency Social (Ministry of Social Social. development) – the survey was developed by the MIDEPLAN Year (survey 1985, 1987, 1990, 1992, Every year since 1990. The survey is 2002, 2004, and 1994, 1996, 1998, 2000, conducted as an additional questionnaire, 2006, 2009 income/wage) 2003, 2006, 2009. 2011 containing 8 questions on earnings from employment and other sources, which is added to the National Employment Survey. Period over Between November and Last quarter of the year (October- which income January December). is assessed Covered Covers the whole population The whole Chilean territory is covered (i.e population including rural areas both urban and rural) with the exception of excluding people who lives areas with difficult access. (This is the areas where access is same coverage as for the national difficult, which represent Employment survey.). People living in only 1.36% of the total collective dwellings (hospitals, prisons, population convents, barracks and others) are excluded. It includes persons aged 15 and over. Sample size 2006: about 73 700 The survey is carried out in 36,000 15 000 households; households (12,000 a month) countrywide, households 2009: about 71 500 through a system of rotating the sections households; and choosing a new selection of homes to 2011: about 59 100 avoid survey fatigue amongst participants households and to keep the sample up to date. Using this system any household included in the survey is interviewed 6 times in an 18 month period. Sample Multistage random Similar procedure to the National procedure methodology. Employment Survey: Two-level stratified A geographic stratification is sample. The method of sampling selection used is based on probability in two stages with geographic stratification by region and by urban-rural area. The values associated with the design are not self- weighted and are adjusted by an exogenous projection of the population calculated by demographic methods agreed to by the INE and CELADE (the Latin American Demographic Center). Response rate Not specified Imputation of yes missing values Other Adjustment is made to Not adjusted with national account. The income adjustments match national accounts distribution is data, by the mean of a factor forced to be for each type of income. equal to the CASEN. Unit for data household and individual The sample design of the survey have collection similar methodological characteristics than the national employment survey in which sampling units are firstly, the sections (residential areas) and secondly, households.

35 National Socioeconomic Encuesta Suplementaria de Ingresos Encuesta de Encuesta Characterisation Survey Proteccion Financiera de (Encuesta CASEN) Social Hogares Break in series The survey is re-designed from 2010: Nueva Encuesta Suplementaria de Ingresos Web source http://www.ministeriodesarro http://www.ine.cl/canales/chile_estadistico/ http://www.pro http://www.bcent llosocial.gob.cl/casen/en/ind mercado_del_trabajo/encuestas_supleme teccionsocial.c ral.cl/estadistica ex.html ntarias/encuestas_suplementarias.php?lan l/index.asp s- g=eng economicas/fina nciera- hogares/index.ht m

The EFH income distribution is forced to be equal to the CASEN figures. As a consequence, the results are the same by construction and this source cannot be considered as a possible alternative to CASEN for measuring income-related variables.

The EPS is not really more frequent than CASEN (it is conducted every second or every third year) and covers less households (15000) than CASEN or ESI.

For these reasons, the remaining of this note will be mostly focused on the comparison between the CASEN survey and the ESI.

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The charts below compare the Gini coefficients and the S80/S20 quintile share ratios from the OECD reference series, based on CASEN1 with data series based on ESI.

Both indicators computed from CASEN are significantly higher than those based on the Encuesta Suplementaria de Ingresos: the Gini index computed from CASEN is about 25 per cent higher than data from the INE survey; the difference is even higher when considering the quintile share ratio -- in the most recent years, the CASEN ratio is nearly the double of the ESI ratio. In each case, the overall trend is broadly similar in both surveys but there are slight differences for particular sub-periods.

According to both data sources, both inequality indicators decreased over the 2000-2006 period. The ESI data suggest that this is mainly due to a fall in 2003/2004. After 2006, both data series suggest broadly a stability in the inequality indicators. For the ESI data, there was a slight fall in 2009 followed by a small regain in 2010 but this should be interpreted with prudence because a redesign of the survey (with the implementation of the so-called NESI) took place by this time.

1 Note that the data shown here refer to a data delivery of revised series in December 2012. These indicators are preliminary and have not been completely validated yet.

36 Figure 8. Trends in Gini coefficients

OECD reference suvery (CASEN) ESI

0.55

0.50

0.45

0.40

Gini (at disposable income) (at disposable Gini 0.35

0.30 1990 1995 2000 2005 2010

Figure 9. S80/S20, Chile

OECD reference suvery (CASEN) ESI

16.0

14.0 12.0 10.0

8.0

S80/S20 6.0 4.0 2.0 0.0 1990 1995 2000 2005 2010

2.1.2 Time series of poverty rates

The Chart below shows the movement over time in the poverty rate according to the CASEN survey. On the basis of the observations available, it appears that this indicator decreased monotonically between 1996 and 2011, by 3 percentage points (from 20.7% to 17.7%). It has unfortunately not been possible to include corresponding data based on the ESI, since, for this survey, the statistical information gathered for this preliminary study (mean income by decile) was not specific enough to compute accurate poverty rates.

37 Figure 10. Trends in poverty rates

OECD reference suvery (CASEN) OECD reference suvery (CASEN)

25.0%

20.0%

15.0% Poverty rate (at 50% median income) median (at 50% rate Poverty 10.0% 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self EmploymentTransfers Taxes Disposable income (HDI) OECD Reference Survey 2009 natcur 1831740 554292 938292 3324324 590532 1500516 123816 -913716 4697569 % av HDI 39.0% 70.8% 12.6% 31.9% 2.6% -19.5% Table 3 lists in detail the income components which have been considered as part of household income in the CASEN data provided to OECD, including a description of transfer programmes covered.

38 Table 3. Income components covered in OECD reference series

Income components

Wage and salary income (EH, ES, EO): - Wage and salaries (excluding employers' contribution to social security) Yes - Related bonuses and commissions Yes - Goods provided by employers No - Severance and termination pay No - Sick paid day paid by the government Yes - Other Please specify

Self-employment income (SE): - Profit/looses from unincorporated enterprise Yes - Net values of goods and services produced for final consumption Yes - Other Please specify Capital income, including private pensions, private occupational pensions and all kinds of private transfers (K): Intereses por depósitos; - Income from financial assets, net of expenses Yes Dividendos por acciones o bonos funancieros; Arriendo de Propiedades Urbanas; Arriendo de Maquinarias, animales o implementos; Arriendo de Propiedades Agrícolas Arriendo de propiedades de temporada - Income from non-financial assets, net of expenses Yes Retiro de Utilidades de empresa - Royalties No Jubilación o pensión de vejez (de origen privado) Pensión de Invalidez (de origen privado) Pensión de Orfandad (de origen privado) - Pensions from individual private plans Yes Montepío o Pensión de Viudez (de origen privado) - Pensions from occupational private plans Yes Seguro de desempleo o cesantía Pensión de Alimentos; Dineros aportados por una persona ajena al hogar - Regular transfers received from/paid to other households Yes Donaciones - Other private transfers Please specify Severance and tarmination pay (if actually unemployed)

Social security transfers from public sources (TR): Please specify programmes names in each case (in national langage and English) Subsidio a la discapacidad mental (mental disability subsidy); Subsidio Familiar Duplo por Invalidez ( Disability subsidy); Pensión Básica Solidaria de Invalidez (Basic solidarity Pension for disability) ; - Accident and disability benefits Yes Aporte Previsional Solidario de Invalidez (Solidarity Previsional Support for Disability) Pensión Básica Solidaria de de Vejez (Basic Solidarity Pension for old age); Aporte Previsional Solidario de Vejez (Previsional solidarity support for old-age); Bono Bodas de Oro (50th wedding aniversary Bonus) Jubilación o pensión de vejez (Pagadas por Dipreca, Capredena o INP/IPS) Pensión de Invalidez (Pagadas por Dipreca, Capredena o INP/IPS) Montepío o Pensión de Viudez (Pagadas por Dipreca, Capredena o INP/IPS); - Old-age cash benefits Yes Pensión por Leye Especiales de Reparación (Pagadas por Dipreca, Capredena o INP/IPS) Subsidio de cesantía (Disability subsidy); - Unemployment benefits Yes Maternal Asistance Subsidy; - Maternity allowance Yes Maternal Asistance Subsidy Subsidio familiar al menor recién nacido (New Born Subsidy); Pensión de Orfandad (Pagadas por Dipreca, Capredena o INP/IPS); Subsidio de Asistencia Familiar (Fmily Asistance Subsidy); Asignación Social (Social Asignment Subsidy); Bono de Protección Familiar (Family Protection Subsidy) ; Bono de Egreso (Graduation Bonus); - Child and/or family allowance Yes Asignación Familiar (Family Asignment). - Housing benefits Yes Bono Invierno (Winter Bonus); Subsidio al Empleo Joven (Young Employment Subsidy); - Other Income-tested and means-tested benefits (please specify) Yes Subsidio de Agua Potable (Drinking Water Subsidy) - Other Please specify

Taxes and social security contributions paid by household (TA) - Income taxes Yes - Taxes on wealth Yes - Employees' social security contributions Yes -Other Please specify Imputation procedures All taxes have been imputed. In this procedure we estimate the amount of tax that the individual should have paid to get the net Please list the above income compononents that have been income reported in the data. To do this we apply the tax paymnet structure reported in Servicio de Impuestos Internos, and imputed and specify the imputation method estimate the gross income (before taxes and social security payment). After this, we apply an effective tax rate for income components wage and salary, Capital and self empleyment income. Finally we apply the social security payment to the estimated income, to get the total tax paid by the persons.

Figure 4 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in GDP, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show pretty similar trends throughout the period.

39 Figure 11. Trends in shares of public social transfers

3 % cash social spending in GDP (OECD SOCX) 2 % transfers in disposable income (OECD reference series)

0 1990 1995 2000 2005 2010

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

One main particularity of the CASEN survey is that micro data from this survey are adjusted to provide income estimates which match the national accounts. This is not the case for any other OECD countries data sources.

The adjustment consists in multiplying the income by (or adding to the income, in the case of property income) a certain factor, which is variable over the years and the types of income. Almost all multiplicative factors are greater than 1, with the exception of those for imputed rents, making the adjusted incomes higher than the unadjusted ones. As increased due to the adjustment is more pronounced in incomes from the highest deciles (i.e. the highest incomes), the Gini coefficient is automatically increased.

According to a June 2011 study2, the variation of the Gini coefficient due to the imputations for adjustment to National Accounts figures was regularly around 7 % for the surveys conducted during the period 1990-2006. This makes the Chilean coefficient less comparable to those of other OECD countries. Nevertheless, the consistency over time in the overestimation leaves the analysis of the movements in the coefficient time series unaffected by the adjustment. Adjustment to National account affects, in theory, other indicators related to income distribution. In particular, it is likely to underestimate the poverty rate and to overestimate the quintile share ratio higher. This adjustment may also explain (at least part of) the gap between CASEN and ESI.

Up to the 2011 data provision, another particularity of the CASEN estimates was that imputed rents as well as income in-kind from self-consumption have been added to cash household income. For the most recent 2012 data provision and modification, these income components have no longer been included in order to match other OECD countries data. The revisions were undertaken for the years 2006, 2009 and 2011.

Another possible reason for differences between Chilean CASEN survey and the OECD framework could be found in the way pension benefits and unemployment and work accidents insurance benefit are

2 Bravo, D. and J. A. Valderrama Torres, “The impact of income adjustments in the CASEN Survey on the measurement of inequality in Chile”, Estudios de Economía. Vol. 38 – No 1, June 2011, pages 43-65. http://www.scielo.cl/pdf/ede/v38n1/art03.pdf

40 classified between public transfers and private transfers/savings/capital income. In the past (up to 2011), all pension benefits have been coded as “public transfers” including the mandatory private pensions which are managed by the private sector. This has been revised in the 2012 update and modifications: In the current Chilean data, pension benefits are treated only as public transfers when they are provided by the state (i.e. the “old system” pension and pensions to the armed forces), while they are considered as capital income when private. Further, the social security contributions represent 20% of the incomes (it is a flat-rate payment in Chile). While all these 20% had been counted as taxes in the former data delivery, a social security tax of only 7% is imputed now. This is the mandatory health insurance. The remaining 13% are not considered since the return of that part of social security payments is treated in the model as capital income/private transfer and not as public transfer. Only those who answer in the survey that have a working contract are assumed to pay 7% for health insurance and are also assumed to pay taxes. Independent workers are assumed to pay taxes only if they mention in the survey that they file for taxes (“boletas” or receipts). Independent workers are assumed to pay 7% health insurance taxes only if they mention in the survey that they pay social security.

Another issue which raised controversial debate in Chilean and international media refers to a modification in the 2011 CASEN survey, with regard to the 2009 survey, namely the inclusion of a category “other income”. In particular, a question was included to get a better idea of the composition of "other incomes" category. It is not easy to assess how it affects the comparison with previous survey, but it concerns a rather restricted part of the respondents: this question were asked only from those having replied 'no' to the question whether they have worked during the week before the survey and 'no' to the question whether they have worked during the month before the survey. Still, the controversy is about whether addition of this component “other income” actually would lead to a lower estimate of the number of poor persons. Chilean representatives from the Ministerio de Desarollo Social and from the Ministry of Finance suggest that a larger part (at least 50%) of this “other income” had actually been recorded in past surveys, and had been hidden in other categories. The Secretariat has requested some sensitivity estimates which include and exclude this income category.

Apart from the methodological changes introduced by this revision, Chilean representatives mentioned an exceptional lump-sum bonus paid in November 2011, while the survey was already 'in the street', and that could not be spread over all months of the year. The effect has been minor.

5. Summary evaluation

Overall, whatever the survey (CASEN or ESI), the few income distribution indicators observed in this review roughly lead to the same conclusions in terms of long-term trends.

However, significant differences between indicators from both surveys in levels are observed, which may be due in particular to the inclusion of the above-described adjustment in the CASEN data. This adjustment may limit the international comparability of the data, although CASEN, based on a big sample, is a very well known and a widely used source of information on income distribution and poverty.

The ESI presents the advantage of being annual and can, thus, allow to capture some short-term movements in the indicators. This latter survey, conducted as an additional questionnaire to the National Employment Survey, could also present consistent statistical characteristics with labour market indicators if they were used in a common study.

41 INCOME DISTRIBUTION DATA REVIEW – CZECH REPUBLIC

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Czech Republic are computed on the basis of two data sets and provided by the Czech National Statistics Office until 2008:

 Indicators for 1992, 1994 and 2002 are based on Micro Census.

 Indicators since 2004 are based on EU-SILC.

According to the Czech data provider, the change in series did not impede on the comparability. Indeed, looking at trends over time, there appears to be no major break in trends.

1.2. National reporting and reporting in other international agencies:

Eurostat’s data series based on annual surveys since 2000.

LIS database, using annual surveys using Micro Census data for years 1992 and EU-SILC for 2004.

The below table presents the main characteristics of those three datasets:

Table 5. Characteristics of datasets used for income reporting, Czech Republic

OECD reference series income Eurostat LIS database distribution database Name EU-SILC EU-SILC Micro Census (1992+1996) EU-SILC (2004) Name of the Eurostat Eurostat Czech National Statistical responsible Office agency Year (survey 2004-2009 2001 (based on underlying 1993, 1997, 2005 surveys and survey) and EU-SILC 2005- representing 1992, 1996, income/wage) 2010 representing income 2004 income years for 2000 and 2004-2009. Period over EU-SILC 2005-2009 Annual income for the all Annual income for the all which income is representing income for year N-1 year N-1 assessed 2004-2009 Covered Household Household population Sample size 7000 dwellings (2009) Actual sample size of 6646 dwellings (2005 11274households. Achieved survey) sample size of 9098 (2010 survey). Sample cross-section, revolving Stratified multi-stage Two-stage design with first procedure panel (annually 25% are sampling stage unit stratification. substituted) Response rate 64.8% (2009) 80.70% 64.80% Imputation of No missing values, negative the missing income was Missing values because of missing values values treated as suggested imputed using correct item non-response as well in the terms of references statistical methods. as partial unit non-response are fully imputed. Unit for data Data are adjusted for annual The primary sample units Household level (for rental collection hours worked to represent (PSUs) are the census income, family/children full-year equivalent earnings. enumeration district (CEUs). allowances, social Earnings are net of The secondary sample units assistance, housing employee social security (SSUs) are the dwelling. allowances, income of contributions but not of persons under 16, transfers income tax between households and income from consumptin of own production), and individual level (for all other incomes).

Break in series Yes No Yes

Web source: http://stats.oecd.org/Index.a http://epp.eurostat.ec.europa http://www.lisdatacenter.org/ spx?QueryId=26068 .eu/portal/page/portal/incom our-data/lis-database/by-

e_social_inclusion_living_co country/czech-republic-2/ nditions/quality/national_qua

lity_reports

43 Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality in the Czech Republic ranks as one of the lowest in the OECD area at 0.254 in 2009 versus 0.314 for the OECD average.

The other series of Gini coefficients on disposable income in Czech Republic are the LIS series and the EU-SILC series. The LIS series shows similarly low levels of income inequality as the OECD series throughout the period 1992 to 2004 (in 2004 with 0.266 for the LIS series and 0.268 for the OECD series). The EU-SILC series is consistent with the OECD series although somewhat lower levels of income inequality than the OECD series throughout.

Figure 12.1 Trends in Gini coefficient (disposable income)

OECD reference series Eurostat (EU-SILC) LIS

0.30

0.25

0.20 Gini (at disposable income) (at disposable Gini

0.15 1990 1995 2000 2005 2010

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference series and the EU-SILC series, the OECD series shows slightly higher levels of income quintile share ratio throughout, although both series show similar trends, with steady levels in the last couple years (3.6 for the OECD series and 3.5 for the EU-SILC series).

44 Figure 1.2 S80/S20

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Czech population living with less than 50% of the median equivalised income (115,000 Korunas per year) has increased only slightly from 5.8% in 2004 to 5.9% in 2009.

While the LIS series is consistent with the OECD reference series, as both show a 5.8% poverty ratio in 2004, the EU-SILC series shows lower levels than the OECD series overall, for instance in 2009 the EU-SILC series shows 5.2% poverty rate levels as opposed to 5.9% poverty rate levels for the OECD series.

Figure 2.1 Trends in poverty rates

OECD reference series Eurostat (EU-SILC) LIS

Poverty rate (at 50% median income) median (at 50% rate Poverty 1%

0% 1990 1995 2000 2005 2010

45 As for child poverty, the OECD reference series and the LIS series both show a large increase since the beginning of the 1990s, reaching 10.3% in 2009 for the OECD reference series, but 7.7% for the EU- SILC series. The trends of the two series are also contrasting as the OECD series shows rising levels of child poverty rates as opposed to the EU-SILC series that shows slightly lower levels.

Figure 2.2 Trends in Child poverty rates

OECD reference series Eurostat (EU-SILC) LIS

12%

10%

2% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 0% 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 6. Shares of income components in total disposable income, OECD reference series

46 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

The OECD reference series (as well as the LIS series) use the square root of household size, whereas the EU-SILC series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14.

Moreover, a comparative table on Living Conditions in Czech Republic in 2011 computed by Jiri Vecernik, researcher at the Academy of Sciences, shows the underestimation of higher earnings in EU- SILC surveys against wage surveys and, therefore, a possible underestimation of income inequality by EU- SILC data. (Source: personal files computed by Jiri Vecernik using wage surveys tables published by the CSO and EU-SILC).

5. Summary evaluation

Currently, there is no other national representative series on household data besides EU-SILC (called Living Conditions survey in the country). All public documents report on EU-SILC data as the only source regarding overall indicators of income inequality and poverty.

The levels and trends of the series are therefore pretty consistent, although the EU-SILC series from Eurostat is generally lower than the OECD series. This could be explained by the different equivalence scales.

It appears that the two different data sources which the OECD reference series are using (Microcensus and EU-SILC) can be compared in levels over time.

47 INCOME DISTRIBUTION DATA REVIEW – DENMARK

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution indicators for Denmark are computed by the Ministry of Finance from their Law Model micro-simulation model. Data are available at the Danish Ministry annually since 1979, but information before 1983 is scarce. Data are published by OECD from 1985 to 2010 annually for Gini and poverty rate.

Note that data from 1995 onwards were revised in 2012 using a larger sample: 1/3 of the population instead of 1/30 previously.

1.2. National reporting and reporting in other international agencies:

Income distribution and poverty indicators for Denmark are also available from:

 Eurostat’s EU-SILC annual survey from 2003.

 Data for 2000 and 2002 are also available from Eurostat based on the EU-SILC predecessor ECHP. Data are not fully comparable.

 LIS database using the Danish Law model in 1987, 1992, 1995, 2000 and 2004

Note that Statistics Denmark’s Statistical Yearbook 2012 reports Ginis from the source Eurostat.

The below table presents the main characteristics of those four datasets:

48 Table 1. Characteristics of dataset, Denmark

49 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the income distribution database, income inequality among the total population has been relatively low and stable at 0.22 between the mid80s and the mid90s. Income inequality then increased between 1995 and 2000 to a Gini of 0.25, before decreasing to 0.23 in 2005. Since 2005, inequality has been fluctuating, with a general upward trend, reaching a Gini of 0.25 in 2010.

Data from the Ministry of Finance from the same data source (Law Model) show a somewhat stronger increase in the Gini, mainly due to the fact that it includes imputed rent and negative capital income (see Annex for a detailed comparison between the OECD reference series and the series from the Ministry of Finance provided by the Danish delegate Lars Pantmann)

By contrast, data from LIS show a decline in inequality between 1987 and 1995, then a regular increase up to 2004, as the other two series based on Law Model data.

Since 2003 from which EU-SILC data are available, the EU-SILC Eurostat series shows a similar upward trend in Ginis than the OECD reference series, with slightly higher levels. Then from 2008 onwards, the Eurostat figure is significantly higher that the OECD and the Danish figures. The difference with the OECD series is still to be determined.

Figure 1.1 Trends in Gini coefficient (disposable income)

Also, when comparing the income quintile ration (S80/S20) from the OECD reference series and the Eurostat EU-SILC, as for the Gini series, the EU-SILC Eurostat series shows a similar upward trend in

50 Ginis with slightly higher levels around 2005. Then from 2008 onwards, the Eurostat figure increases to 4.5, compared to 3.5 for the OECD figure. The difference with the OECD series is still to be determined.

Figure 1.2 Trends in S80/S20

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Danish population living with less that 50% of the annual median equivalised income (111 757 DKK in 2010) has been fluctuating at around 6% in the last decades. The relative poverty rate decreased from around 7.5% in 1987 to around 5% in 1995-2000. It then increased from 5.6% in 2005 to 6.6% in 2008, before declining again in the last years to 6.0% in 2010.

Data from the Ministry of Finance from the same source (Law Model) shows lower poverty rates, mainly due to the fact that it includes imputed rent and negative capital income (see Annex for a detail comparison between the OECD reference series and the series from the Ministry of Finance provided by the Danish delegate Lars Pantmann). OECD figures do not include as income negative capital income and imputed rent from private housing, which gives an incomplete picture of the income situation, in particular for older people. When an imputed rent is included in the definition of income, the risk of poverty in Denmark for elderly people is much lower.

The LIS series shows higher poverty rates around 1990 but it shows similar rates to the OECD series between the mid90s and the mid2000s. Without taking into account the 4.0% from ECHP in 2000, the EU- SILC series is very similar to the OECD series, again up to 2008 where the EU-SILC series becomes higher. The difference with the OECD series is still to be determined.

51 Figure 2.1 Trends in poverty rates

As for child poverty rates, the three series available show similar trends, but the LIS series (except in 2000) and, in particular, the EU-SILC series show rates which are 1 to 2 percentage points higher.

Figure 2.2 Trends in poverty rates among children

2.2 Wages

See Part II of the present Quality Review.

52 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

When comparing the composition of the average equivalised disposable income of the OECD reference series (based from the Law model, DKK) and the EU-SILC series (in Euros), share of income generally match, except for the share of capital (13% in the OECD series, 6% in the EU-SILC series) and that of transfers (22% in the OECD series, 31% in the EU-SILC series).

The shares of taxes are surprisingly similar, as the OECD series would not include social security contributions in taxes.

Table 2. Shares of income components in total disposable income, OECD reference series and EU SILC

Figure 3. Trends in shares of public social transfers

53 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Data from the Ministry of Finance from the same source (Law Model) includes imputed rent and negative capital income (see Annex for a detail comparison between the OECD reference series and the series from the Ministry of Finance provided by the Danish delegate Lars Pantmann).

The difference between the OECD series and EU-SILC series from 2008 is still to be determined.

5. Summary evaluation

Overall the OECD series matches reasonably well with the Eurostat series up to 2007. From 2008 onwards, the Eurostat series shows higher levels of both Ginis and poverty rates.

The LIS series also show comparable figures. Unfortunately it is only available until 2004.

Figures from the Law model are now calculated based on a sample accounting for one-third of the Danish population, which minimizes sample errors and gives higher credibility to the figures.

54 Annex: Methodological Note from Lars Pantmann, Delegate from Ministry of Finance, comparing Danish National figures and OECD figures

OECD Income distribution – Compared to National figures

Figures supplied to the OECD Income distribution study are calculated on the same data source as used in official national income distribution analyses, i.e. The Danish Law model system3. However there are a number of methodological differences in the calculation of income.

The following tables summaries some of the resulting differences such as Gini and the share of the population with relative low income (lower than 50 percent of the median income).

Table 1.1

Gini – total population

1985 1990 1995 2000 2005 2010

OECD 22.09 22.56 21.49 22.72 23.19 25.21 National 19.61 20.56 19.92 22.41 23.28 25.66

Difference -2.48 -2.00 -1.57 -0.31 0.09 0.45 Due to Including negative income elements1) -2.15 -1.47 -1.64 -0.76 -0.90 0.37 Including an imputed rent element -0.25 -0.33 0.12 0.36 0.99 0.40 Including single person households under 18 0.09 0.07 0.07 0.08 0.08 0.08 Equvalence scale (0,5  0,6) -0.32 -0.36 -0.38 -0.31 -0.31 -0.43 Housing unit  Families 0.29 0.41 0.42 0.54 0.54 0.74 Top-/bottom coding -0.13 -0.33 -0.17 -0.23 -0.31 -0.71

Note: The decomposition of the changes is influenced by the sequence of calculation. In OECD study the persons sharing the same housing unit is regarded as the income sharing unit. In national studies the housing unit can contain more income sharing units (families). A consequence of that is, that children living in another housing unit than there parents have to be excluded in national studies. 1) In OECD calculations only the positive part of each of the elements Salary income, self employed income, capital income, transfers and taxes are included in the disposable income. Source: Calculations based on a 33.3 percent sample of the Danish population.

Including negative income elements an imputed rent element has significant effect on the interpretation of the analysis:

 The trend – negative net capital income is reduced markedly from 1985 to 2005. This reduction leads to at rise in the national Gini witch isn’t reflected in the OECD Gini.

 The composition of the low-income-group – negative net capital income lowers the median income and thereby the low-income-threshold. Imputed rent increases the income of the retirement age population (they typically don’t have negative capital income any longer). In

3 The Danish Law model system includes very detailed information about 33.3 percent of the population.

55 national studies about 1 percent of the retirement age population has relative low income where it’s about 10-20 percent in OECD studies, see table 2.3.

Negative capital income reduces the income tax. Not including negative capital income therefore introduces a mismatch between income and taxes. A similar situation applies to negative self employed income.

In order to compare income of house owners with tenants an imputed rent of 4 percent of the property valuation is added to the capital income for owners in national studies. The imputed rent has contributed to an increase in the national calculated Gini.

Even though the sample used is large, a few very extreme observations can introduce some noise when the changes over time is analysed. Therefore a disposable income over 20 million Danish kroner or below minus 5 million Danish kroner is excluded from the sample in the national studies.

Looking at the income of those with these very high disposable incomes reveals that the high income is due to equity income or self employed income. It is very likely that these incomes relates to a period longer than one year. Ideally the income should be spread over the total period and not only in the payment year. However that is a serious task – partly because of tax payment.

Table 1.2

Gini – working age population

1985 1990 1995 2000 2005 2010

OECD 20.89 21.49 20.60 21.90 22.71 24.84 National 19.32 20.73 20.06 22.57 23.58 26.24

Difference -1.56 -0.76 -0.54 0.66 0.87 1.40 Due to Including negative income elements1) -1.58 -0.86 -1.02 -0.17 -0.44 0.66 Including an imputed rent element -0.25 -0.31 0.15 0.41 1.07 0.64 Including single person households under 18 0.00 0.00 0.00 0.00 0.00 0.00 Equvalence scale (0,5  0,6) -0.22 -0.30 -0.30 -0.23 -0.24 -0.35 Housing unit  Families 0.67 0.85 0.85 0.94 0.93 1.23 Top-/bottom coding -0.18 -0.13 -0.21 -0.29 -0.45 -0.78

Note: See note in table 1.1. 1) In OECD calculations only the positive part of each of the elements Salary income, self employed income, capital income, transfers and taxes are included in the disposable income. Source: Calculations based on a 33.3 percent sample of the Danish population.

56 Table 1.3

Gini – Retirement age population

1985 1990 1995 2000 2005 2010

OECD 19.67 20.06 18.83 20.70 20.55 21.94 National 17.85 18.42 18.13 20.80 21.70 22.28

Difference -1.83 -1.63 -0.70 0.11 1.15 0.34 Due to Including negative income elements1) -0.26 1.19 0.13 0.37 0.14 1.12 Including an imputed rent element -0.16 -0.37 0.06 0.59 1.60 0.79 Including single person households under 18 0.00 0.00 0.00 0.00 0.00 0.00 Equvalence scale (0,5  0,6) -0.70 -0.63 -0.70 -0.68 -0.66 -0.68 Housing unit  Families -0.65 0.09 -0.05 0.11 0.15 0.14 Top-/bottom coding -0.05 -1.91 -0.13 -0.28 -0.08 -1.02

Note: See note in table 1.1 1) In OECD calculations only the positive part of each of the elements Salary income, self employed income, capital income, transfers and taxes are included in the disposable income. Source: Calculations based on a 33.3 percent sample of the Danish population.

Table 2.1

Share of the population with relative low income – total population

1985 1990 1995 2000 2005 2010

OECD 6.00 6.16 4.73 5.12 5.32 6.00 National 3.92 4.29 3.30 3.98 4.94 6.19

Difference -2.08 -1.87 -1.42 -1.15 -0.38 0.19 Due to Including negative income elements1) -2.30 -1.83 -1.43 -1.40 -0.83 -0.40 Including an imputed rent element -0.23 -0.44 -0.04 0.15 0.23 0.18 Including single person households under 18 0.11 0.09 0.09 0.09 0.10 0.10 Equvalence scale (0,5  0,6) -0.13 -0.25 -0.53 -0.60 -0.44 -0.34 Housing unit  Families 0.56 0.64 0.58 0.73 0.72 1.00 Top-/bottom coding -0.09 -0.09 -0.10 -0.11 -0.16 -0.35

57 Table 2.2

Share of the population with relative low income – working age population

1985 1990 1995 2000 2005 2010

OECD 4.12 4.92 3.95 4.41 5.12 6.30 National 4.38 5.05 4.05 4.91 6.06 7.70

Difference 0.26 0.13 0.10 0.50 0.94 1.40 Due to Including negative income elements1) -0.33 -0.22 -0.26 -0.29 -0.09 0.17 Including an imputed rent element -0.07 -0.26 0.05 0.30 0.55 0.59 Including single person households under 18 -0.01 -0.00 -0.00 -0.01 -0.01 -0.01 Equvalence scale (0,5  0,6) -0.27 -0.45 -0.56 -0.57 -0.55 -0.53 Housing unit  Families 1.06 1.17 1.01 1.22 1.27 1.74 Top-/bottom coding -0.13 -0.11 -0.14 -0.15 -0.24 -0.55

Table 2.3

Share of the population with relative low income – retirement age population

1985 1990 1995 2000 2005 2010

OECD 19.46 16.56 12.29 12.74 10.12 7.97 National 1.25 1.29 0.78 0.96 1.15 1.22

Difference -18.21 -15.27 -11.50 -11.79 -8.98 -6.74 Due to Including negative income elements1) -16.52 -13.70 -9.63 -9.09 -5.79 -3.76 Including an imputed rent element -1.09 -1.14 -0.19 -0.30 -1.35 -1.86 Including single person households under 18 -0.01 0.00 -0.02 -0.01 -0.01 -0.01 Equvalence scale (0,5  0,6) -0.48 -0.40 -1.68 -2.52 -1.91 -1.24 Housing unit  Families -0.05 0.08 0.09 0.22 0.16 0.19 Top-/bottom coding -0.06 -0.11 -0.07 -0.09 -0.08 -0.06

58 INCOME DISTRIBUTION DATA REVIEW – ESTONIA

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Estonia are computed from the EU-SILC survey from the 2004 onwards. Until recently, the indicators have been provided directly by Statistics Estonia.

1.2. National reporting and reporting in other international agencies:

 Estonian Social Survey (ESS) using annual data from ‘Statistics Estonia’ since 2004.

 Eurostat’s EU-SILC annual survey since 2000.

 LIS database, using surveys from the ‘Estonian Social Survey’ for 2000 and 2005.

The below table presents the main characteristics of those four datasets:

59 Table 7. Characteristics of datasets used for income reporting, Estonia

60 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among total population has decreased overall in Estonia in contrast to most other OECD countries, which have seen a steady increase. The OECD reference series shows a general decline from 0.349 in 2004 to 0.314 in 2009, although the trend has stabilized since 2007 (0.313).

From the three other series of Gini coefficients on disposable income in Estonia, the EU-SILC series and ESS series are almost identical, although the EU-SILC series covers a longer period with an unusually large increase in 2003 (0.374). All series seem to be converging since 2008. The LIS database, although limited to 2000 and 2005 shows a decline that is line with the general pattern.

Figure 13.1 Trends in Gini coefficient (disposable income)

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference survey and the Eurostat EU-SILC, both series point to a general decline and reach similar levels in 2009 with a ratio of 5.1 for the OECD series and of 5 for the EU-SILC series.

61 Figure 1.14 S80/S20

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Estonian population living with less than 50% of the median equivalised income (3970 Euros per year in 2008) has declined from 14.2% in 2004 to 11.1% in 2009. The EU-SILC series, although not as significant, also shows a decline in poverty rates since 2007, reaching 9.4% in 2009.

Figure 2.1 Trends in poverty rates (50% median)

As for child poverty, the OECD reference series and EU-SILC series show a decline, with OECD series suggesting a stronger decline, both series reaching 12.1% in 2008. By contrast, the LIS series, although limited to 2000/2005 shows an increase in child poverty rates. However, this is difficult to compare since the LIS series covers a period that does not overlap with the two other series.

62 Figure 2.2 Trends in Child poverty rates

2.2 Wages

See Part II of the present Quality Review

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 8. Shares of income components in total disposable income, OECD reference series

63 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and the other income series:

The OECD reference series (as well as the LIS series) use the square root of household size, whereas the EU-SILC series and ESS series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

While the levels and trends of the series are generally quite similar overall, the EU-SILC series gives frequently somewhat lower inequality and poverty estimates than the OECD reference series. This may be due to the different equivalence scales used (square foot of household size for the OECD series; OECD modified equivalence scale for EU-SILC).

64 INCOME DISTRIBUTION DATA REVIEW – FINLAND

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Finland are provided by Statistics Finland and are based on different surveys:

 Data for 1976 are based on the Household Budget Survey (HBS)

 Data for 1986, 1995 and 2000 are based on the Income Distribution Survey (IDS)

 SData for 2004, 2008, 2009 and 2010 are based on the Income and Living Conditions (previously labelled as the Income Distribution Survey), which is the Finnish version of EU-SILC.

According to Statistics Finland these data series are comparable over time.

1.2. National reporting and reporting in other international agencies:

 Statistics Finland conducts the Income Distribution Statistics survey annually since 1977. 4

 EUROSTAT has been computing indicators on inequalities and poverty for Finland from 1996 (income year 1995). The primary sources of data are European Community Household Panel (ECHP) 1995-2000 and European Statistics on Income and Living Conditions (EU-SILC) 2003 to present. In 2001 and 2002, data was provided by Statistics Finland from national sources. 5

 Finland has been included in the EU-SILC (Statistics on Income and Living Conditions) survey since 2004 (income year 2003). EU-SILC is a multi-dimensional instrument focused on the income and the living conditions of different types of households. It is collecting, on an annual basis, timely and comparable multidimensional micro-data on income, material deprivation, housing condition, labour, education, health and subjective well-being.

 The Luxembourg Income Study Database (LIS) included Finland in years 1987, 1991, 1995, 2000, and 2004. It is based on the Income Distribution Survey that is presented in more details in the below table. 6

The below table presents the main characteristics of those four datasets:

4. See http://www.stat.fi/meta/til/tjt_en.html 5. See http://epp.eurostat.ec.europa.eu/cache/ITY_SDDS/EN/ilc_esms.htm 6 . See http://www.lisdatacenter.org/our-data/lis-database/by-country/finland-2/

65 Table 9. Characteristics of datasets used for income reporting, Finland

OECD reference series Statistics Finland LIS database Eurostat income distribution database HBS 1976 Income Distribution Income Distribution ECHP: 1995-2000 Name IDS 1986 1995 200 Statistics; Income Survey(Tulonjakotilasto) IDS: 2000-2001 IEU-SILC 2004, 2008 Distribution Survey EU-SILC: 2002-present and 2009. Name of the HBS: Statistics Finland Statistics Finland Statistics Finland EUROSTAT responsible agency IDS: Statistics Finland EU-SILC: Eurostat Year (survey and 1976, 1986, 1995, 1966-2010 1987, 1991, 1995, 2000, 1996-2011 surveys income/wage) 2000, 2004, 2008, 2004 representing income for 2009 years 1995-2010.

Period over which Annual years Annual year Annual year Annual year N-1 income is assessed

Covered population Permanent household Permanent household Permanent household in Permanent household in Finland (overseas in Finland (overseas Finland (overseas in Finland (overseas dwellers, those without dwellers, those without dwellers, those without dwellers, those without address, institutional address, institutional address, institutional address, institutional population is population is population is excluded). population is excluded). excluded). Same excluded). Same Same households are Same households are households are households are included in 2~4 included in 2~4 included in 2~4 included in 2~4 consecutive years. consecutive years. consecutive years. consecutive years. Sample size 13524 households About 10,000 2004: 11 229 households 13524 households households containing 29112 individuals

Sample procedure cross-sectional sample survey and Two-phase stratified PPS Two-phase stratified interviews. design/Administrative sampling based on a registers complied with nationally interviews. representative probability sample of the population residing in private households.Sampling unit is a person. In the first phase persons are selected (target persons), in the second phase the target persons together with their household- dwelling units are selected. Response rate EU-SILC: 92% 85.20% 92% Imputation of missing No missing values, Yes Yes values negative values treated as suggested in the terms of references Unit for data household household household Household collection

66 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below figure shows the evolution of Gini coefficients for Finland from 1985 to 2010, as reported by the OECD, LIS, Statistics Finland and the EU-SILC.

Figure 15.1 Trends in Gini coefficient (disposable income)

OECD reference series LIS (IDS) Statistics Finland (IDS) Eurostat (EU-SILC)

0.30

0.28

0.26

0.24

0.22 Gini (at disposable income) (at disposable Gini 0.20

0.18 1985 1990 1995 2000 2005 2010

According to the OECD series, income inequality in Finland rose over the last two decades, from 0.209 in 1987 to 0.259 in 2009. However, the level is still largely below the OECD average of 0.314, according to the OECD database.

The trend of the OECD series is similar to the other series, while it shows slightly lower levels of income inequality overall. Indeed, Statistics Finland shows higher levels of income inequality throughout the year 2000s, reaching 0.266 points in 2009. Although generally higher than the OECD series, the Eurostat series is almost identical since 2008, exhibiting 0.254 in 2009. Finally, the LIS series shows levels and trends similar to the OECD series, although slightly higher for the year 2000.

Also, when comparing the income quintile share ratio (S80/S20) from the OECD series with the series from the Eurostat and Statistics Finland, the trends and levels are generally quite similar and also show an increase since the 1990s, reaching 3.7 in 2009 for the OECD series, 3.8 points in 2010 for the Statistics Finland series, and 3.7 in 2010 for the EU-SILC series.

67 Figure 1.2 S80/S20

OECD reference series Statistics Finland (IDS) Eurostat (EU-SILC)

5.0

4.5

4.0

3.5 S80/S20

3.0

2.5

2.0 1990 1995 2000 2005 2010

2.1.2 Time series of poverty rates

Poverty rates for Finland all show a pretty steady increase for the period 1985 to 2010. However, there is a visible decline in the early 1990s as well as in the year 2009 for the OECD, EU-SILC and Statistics Finland series. Thus, poverty rates (at 50% median income) in Finland concern 7.3% of the total population in 2009 according to the OECD series, 6.5% in 2010 according to Statistics Finland, and 6% in 2010 according to EU-SILC.

The OECD series shows levels that are almost constantly higher than EU-SILC and Statistics Finland, yet the trends are very similar. The LIS series is similar to the OECD series for the period covered by both series.

Figure 2.1 Trends in poverty rates

OECD reference series LIS (IDS) Statistics Finland (IDS) Eurostat (EU-SILC) 10%

2% Poverty rate (at 50% median income) median (at 50% rate Poverty

0% 1985 1990 1995 2000 2005 2010

68 As for child poverty, the trends are similar to the general poverty rates, with a substantial rise throughout the entire period, before a sizeable drop since 2008. Thus, child poverty rates (at 50% median income) in Finland are of 4.4% in 2009 according to the OECD series, 5.3% in 2010 according to Statistics Finland, and 4.1% according to EU-SILC. The Statistics Finland series thus exhibits generally higher levels than the other series.

Figure 2.2 Trends in Child poverty rates

OECD reference series LIS (IDS) Statistics Finland (IDS) Eurostat (EU-SILC)

10%

2% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 0% 1985 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Unfortunately, such information could not be found for alternative data sources.

Table 10. Shares of income components in total disposable income, OECD reference series

69 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and other income series:

Equivalence scale: The OECD reference series, as well as the LIS series, use the square root of household size, whereas the EU-SILC series and Statistics Norway series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

Occupational pensions: Until recently, the Finnish occupation pension benefits have been recorded in the OECD database framework as “private transfers/savings/capital income”, rather than as part of public social transfers. This has been revised in the 2011 update, after internal discussion and discussion with Finnish authorities. The revision was undertaken backwards (except for the year 1976), and Finnish occupational pensions are now classified as public transfers.

5. Summary evaluation

Broadly speaking, the different indicators for Finland follow similar trends throughout the covered period. Income inequality and poverty rates have increased ubiquitously across the different series. This being said, levels of both inequality and poverty indicators have slightly dropped since 2008, though there might be an upswing again in 2010. Also, it is visible that the Statistics Finland series shows generally higher levels than the other series, with the exception of poverty rates where levels shown by the OECD series are higher than the other series.

70 INCOME DISTRIBUTION DATA REVIEW - FRANCE

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for France are computed by INSEE from the annual Enquête Revenus fiscaux et sociaux (ERFS). ERFS data are available from 1996. Data before 1996 are from the annual Enquête Revenus Fiscaux (ERF) and are available for the years 1984 and 1996 in the OECD database; those are, however, considered to have lower quality data on social benefits and capital income and are therefore only partially comparable with the later data series.

1.2. National reporting and reporting in other international agencies:

Income distribution and poverty indicators for France are also available from

 National ERFS series published annually by INSEE

 Eurostat’s EU-SILC annual survey since 2005, and Eurostat’s ECHP annual survey, available from 2000 to 2004.

 LIS database, using the French Household Budget Survey from INSEE (Enquête Budget de famille) in 1984, 1989, 1995, 2000 and 2005

The below Table 1 presents the main characteristics of those three datasets:

71 Table 1. Characteristics of dataset used for income reporting, France

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among total population has remained broadly stable in France since 1990, in contrast to most other OECD countries which have seen a steady increase. From a modest decline in the 1990s, inequality in France has been slowly increasing during the 2000s, since 2005.

The OECD reference Gini series is very similar in level and trend to the French INSEE series published at national level – they are both based on the same ERFS survey. The EU-SILC series from Eurostat shows a similar upward trend in the 2000s with similar Ginis in 1999, 2003 and 2007. But the Eurostat series shows more variation during intermediate years. Finally, overall the LIS series shows a similar stable trend, but data are only available until 2005.

72 Figure 1.1. Gini coefficients, France

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference survey and the Eurostat EU-SILC, as for the Gini series, the EU-SILC series from Eurostat shows a similar upward trend in the 2000s and particularly between 2009 and 2010, with similar Ginis in 1999, 2003 and 2007. But the Eurostat series shows more variation during intermediate years.

Figure 1.2. S80/S20, France

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the French population living with less than 50% of the median equivalised income (10 406 Euros per year in 2010) has remained broadly stable at around 7% until 2008. It recently increased to 7.5% in 2009 and 7.9% in 2010.

73 The OECD reference series is very similar to the French INSEE series – they are both based on the same ERFS survey. However, the EU-SILC series from Eurostat and the LIS eries based on HBS shows a different downward trend in the late 1990s, from 9% in 1994 to 6% in 6% in 2001-02 for the former. The LIS series shows a different upward trend in relative poverty from 7.3% 2000 to 8.5% 2005.

Figure 2.1 Trends in poverty rates

OECD reference survey EU-SILC Eurostat INSEE LIS (HBS) 12.0%

10.0%

8.0%

6.0%

4.0%

2.0% Poverty rate (at 50% median income) median (at 50% rate Poverty

0.0% 1990 1995 2000 2005 2010

As for child poverty, the OECD reference series shows a continuous upward trend. The INSEE series shows a different downward trend from 10.3% in 1996 to 9.3% in 1999, then it shows some variation in 2002 (8.0%) and 2004 (8.4%). It looks similar to the OECD series since the revised ERFS from 2005 onwards. Although rates are different, other sources also show an upward trend in poverty rates among children, particularly with a steeper increase from 2005 on EU-SILC series.

Figure 2.2. Child Poverty rates, France

OECD reference survey EU-SILC Eurostat INSEE LIS (HBS) 12.0%

10.0%

8.0%

6.0%

4.0%

2.0% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 0.0% 1990 1995 2000 2005 2010

74 As for child poverty, the OECD reference series shows a continuous upward trend. The INSEE series shows a different downward trend from 10.3% in 1996 to 9.3% in 1999, then it shows some variation in 2002 (8.0%) and 2004 (8.4%). It looks similar to the OECD series since the revised ERFS from 2005 onwards. Although rates are different, other sources also show an upward trend in poverty rate among children, particularly with a steeper increase from 2005 on EU-SILC series.

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. When comparing the composition of the average equivalised disposable income of the OECD reference survey (based from ERFS) with the EU-SILC series, shares of income generally match, except for the share of taxes. The latter are estimated to be much lower in ERFS. This is due to the fact that the ERFS excludes some social contributions, which underestimates the effect of taxes (14% of HDI against 23% using EU-SILC).

Table 2. Main income aggregates, France, 2008

75 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD Terms of References and the methodology used by INSEE on ERFS:

Children are defined as non-married children in the household, without limit of age, rather than persons below age 18, as suggested in the OECD Terms of Reference

The work status of persons is taken from data matching with the LFS and the definitions therefore are conform to LFS concepts, rather than defining “work” as having non-zero earnings as suggested by the OECD Terms of Reference)

Methodological differences between the OECD reference series based from ERFS and French results based on ERFS:

Equivalence scale: the OECD reference series (as well as the LIS series) uses the square root of household size (so does the LIS series), whereas the INSEE-ERFS series (as well as the Eurostat series) uses the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

The OECD reference series and the INSEE series (coming from the same survey) show very similar results. The OECD reference series based from INSEE-ERFS show less variation (are smoother) than EU- SILC series. This is probably due to the fact that ERFS is based on a larger sample than EU-SILC (37 000 > 13 500 households). There are, however, more differences in estimates between these series for the period of the 1990s and early 2000s, than since the early 2000s.

The main issue arising from the comparison of ERFS with EU-SILC data arises from differences in covering income taxes, in particular social security contributions. It seems that the INSEE-ERFS series therefore underestimate the volume of income taxes. This is particularly important for estimates of redistribution.

Bibliography:

Insee (2012), Revenus-Salaires - Les niveaux de vie en 2010 (www.insee.fr/fr/themes/document.asp?ref_id=ip1412)

76 INCOME DISTRIBUTION DATA REVIEW –GERMANY

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD Income Distribution Data for Germany has been using data from the German Socio- Economic Panel Study (GSOEP) provided by the Deutsches Institut für Wirtschaftsforschung (DIW Berlin). Data available in the OECD database refer to the years 1985, 1990, 1995, 2000, 2008, 2008, 2009 and 2010. Data points for 1985 and 1990 refer to the old German Länder, data points for 1995 and thereafter to both new and old Länder. Summary data for 1985 and 1990 have been adjusted backwards to estimate an overall distribution, applying a splicing with the two available data points for 1995.

The German Socio-Economic Panel Study (SOEP) is a wide-ranging representative longitudinal study of private households, located at the German Institute for Economic Research, DIW Berlin. Every year, there were nearly 11,000 households, and more than 20,000 persons sampled by the fieldwork organization TNS Infratest Sozialforschung. The data provide information on all household members, consisting of Germans living in the Old and New German States, Foreigners, and recent Immigrants to Germany. The Panel was started in 1984. Some of the many topics include household composition, occupational biographies, employment, earnings, health and satisfaction indicators.

In 2009/10, GSOEP data provided by DIW have been revised backward till 1984 in order to take into account partial non response on the composition of income. Missing income data have been imputed. This change has led to an increase in the level of income and mostly for families with children therefore the poverty rate estimates for children have decreased.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Income distribution and poverty indicators for Germany are also available from:

 Einkommens- und Verbrauchsstichprobe (EVS): The Income and expenditure survey EVS is a large and detailed survey on income, expenditure and assets. The sample survey of income and expenditure (German abbreviation: EVS) provides important official statistics on the standards of living of households in Germany. Among other things, it provides statistical information on the households' equipment with consumer durables, on their income, property and debt situation as well as on their final consumption expenditure. The sample survey of income and expenditure is conducted at five-year intervals. Since there is no legal obligation for households to participate in the survey, any information is rendered on a voluntary basis.

 Microcensus: The microcensus provides official representative statistics of the population and the labour market in Germany. The Labour Force Survey of the European Union (EU Labour Force Survey) forms an integral part of the microcensus. Since 1957 - in the new Länder (including Berlin-East) since 1991 - the microcensus has supplied statistical information in a detailed subject-related and regional breakdown on the population structure, the economic and social situation of the population, families, consensual unions and households, on employment, job search, education/training and continuing education/training, the housing situation and health.

77  Continuous household budget surveys (LWR): In the context of the continuous household budget survey, households in Germany are requested to provide annual information on household income and expenditure, housing conditions and possession of consumer durables. The current concept of the continuous household budget surveys that has been in use since 2005 is a sample survey covering 8,000 households all over Germany on an annual basis.

1.2.2 International reporting:

 Eurostat is also computing and publishingindicators on income distribution and poverty for Germany based on EU-SILC.

 Germany is also included in the Luxembourg Income Study Database (LIS) using the German Social Economic Panel Study (GSOEP) since 1984 and the Income and Consumer Survey (EVS) before.

Table 1. presents the main characteristics of the different sources:

Table 1. Characteristics of datasets used for income reporting, Germany

78 Table 1. Characteristics of datasets used for income reporting, Germany (cont.)

79 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

GSOEP data provided by DIW have been revised in 2009/10 backward till 1984 in order to take into account of partial non response on the composition of income. Missing income data have been imputed. This change has led to an increase in the level of income and mostly for families with children therefore the poverty rate for children has decreased. As LIS has not revised their data to account of partial non response, hence the data are not fully comparable to the OECD.

In the past, EU-SILC and GSOEP used to report different levels of inequality and poverty but have tended to converge. This may be linked to the fact that EU-SILC tended to undercover the immigrant population in the past which has resulted of an overestimation of income level and lower inequality.

OECD and (national) GSOEPv28 times series are very close and may differ because of a different equivalence scale. OECD uses the scare root definition whereas DIW (GSOPE) uses the so called “modified OECD scale:” where the head of household is assigned a weight of 1, children up to the age of 14 a weight of 0.3 and all other household members a weight of 0.5.

Figure 1.1 Trends in Gini coefficient (disposable income), Germany

Source: German Socio-Economic Panel Study GSOEP (DIW Berlin); Income and expenditure survey (EVS); Continuous household budget surveys (LWR); Mikrozensus; Eurostat; EU Survey on Income and Living Conditions (EU SILC), European Community Household Panel (ECHP), LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

80 2.1.2 Time series of poverty rates

Figure 2.1 Poverty rates, 50% threshold, after tax and transfers, Germany

Source: German Socio-Economic Panel Study GSOEP (DIW Berlin); Eurostat; EU Survey on Income and Living Conditions (EU SILC), European Community Household Panel (ECHP), LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

Figure 2.2 Child Poverty rates, 50% threshold, after tax and transfers, Germany

Poverty rates based on OECD and LIS differ due to the revision on the imputation of missing component of income as LIS has not revised their data to account for partial non response in income component. OECD and Eurostat estimates of poverty rates differ more largely.

81 2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for 2008, according to the OECD benchmark series (GSOEP) and according to EU-SILC. It can be seen that EU-SILC data reports a lower share of capital and in particular self-employment income with regard to the OECD benchmark series.

Table 2. Shares of income components in total disposable income, OECD reference series and EU-SILC

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference series (GSOEP) 2008 natcur 10 069 3 911 3 216 17 196 1 989 5 212 5 926 -8 049 22 276 % av HDI 45% 77% 9% 23% 27% -36%

EU SILC 2008 natcur 14 465 5 498 6 19 969 1 200 2 229 6 924 -7 745 22 758 % av HDI 64% 88% 5% 10% 30% -34%

natcur 0.70 0.86 1.66 2.34 0.86 1.04 0.98 % av HDI 0.71 0.88 1.69 2.39 0.87 1.06 Table 3 below provides information on the definition and classification of income components including government transfers, provided by DIW (The German Institute for Economic Research).

82 Table 3. Definition and classification of income components in OECD benchmark series

Income components Wage and salary income (EH, ES, EO): - Wage and salaries (excluding employers' contribution to social security) Yes - Related bonuses and commissions Yes - Goods provided by employers No - Severance and termination pay Yes - Sick paid day paid by the government No - Other Self-employment income (SE): - Profit/looses from unincorporated enterprise Yes - Net values of goods and services produced for final consumption No - Other Capital income, including private pensions, private occupational pensions and all kinds of private transfers (K): - Income from financial assets, net of expenses Yes - Income from non-financial assets, net of expenses Yes - Royalties No - Pensions from individual private plans Yes - Pensions from occupational private plans Yes - Regular transfers received from/paid to other households Yes - Other private transfers

Social security transfers from public sources (TR): Yes (Gesetzliche Unfallversicherung = accident insurance, Gesetzliche - Accident and disability benefits Pflegeversicherung = nursing care insurance)

Yes (Gesetzliche Rentenversicherung = statutory pension system) - Old-age cash benefits

Yes (Arbeitslosengeld 1 = unemployment benefit 1) - Unemployment benefits Yes (Mutterschaftsgeld, Erziehungsgeld, Elterngeld = maternity allowance, - Maternity allowance child raising allowance, parents money)

Yes (Kindergeld = child benefit) - Child and/or family allowance Yes ( Wohngeld = housing allowance, Wohneigentumsförderung = home - Housing benefits buyer allowance) Yes (Arbeitslosengeld 2 = unemployment assistance 2, Sozialhilfe = social assistance, Grundsicherung im Alter = social assitance for the elderly, - Other Income-tested and means-tested benefits (please specify) Kinderzuschlag = childrens allowance) BAFOEG, Stipendien = Student grants, Unterhaltsvorschusskasse = Guaranteed/Advance Maintenance Programs - Other Taxes and social security contributions paid by household (TA) - Income taxes Yes - Taxes on wealth No (only relevant for owner occupiers but a rather small amount) - Employees' social security contributions Yes -Other Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show broad similar trends throughout the period, even though the decline between 2006 and 2008 seems to be slightly more pronounced in the OECD Social Expenditure database (SOCX).

83 Figure 3 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD Terms of References and the methodology used by DIW on GSOEP: OECD uses the equivalence scale based on the square root of the household size whereas DIW uses the old OECD equivalence scale. Some of the nationally published series also include imputed rent as income component which is excluded from the OECD data.

As mentioned earlier, in 2009/10, GSOEP data provided by DIW have been revised backward till 1984 in order to take into account of partial non response on the composition of income and missing income data have been imputed. This change has led to an increase in the level of income and mostly for families with children therefore the poverty rate for children has decreased.

There are, however, larger differences in estimates between the OECD series and the Eurostat series based on EU-SILC, especially in the year 2004 - 2006.

5. Summary evaluation

The German Socio-Economic Panel Study (GSOEP) seems to be the most appropriate source for income distribution and poverty data in Germany due to the quality of the data. Differences of estimates which existed with data based on EU-SILC in the past tended to become smaller in the past years.

Bibliography:

DIW Wochenbericht,Wirtschaft. Politik Wissenschaft. Seit 1928, 2012, Einkommensentwicklung und Armutsrisiko.

SOEP papers on Multidisciplinary Panel Data Research Income Inequality and Poverty in Front of and During the Economic Crisis – An Empirical Investigation for Germany 2002-2010, Jürgen Faik

84 Dealing with incomplete household panel data in inequality research, SOEP papers on Multidisciplinary Panel Data Research, Berlin, March 2010, Joachim R. Frick • Markus M. Grabka • Olaf Groh-Samberg

SOEP monitor 1984-2010 Zeitreihen zur Entwicklung ausgewählter Indikatoren zu zentralen Lebensbereichen/ Time Series on selected Indicators about “Living in Germany”

SOEP papers on Multidisciplinary Panel Data Research, The evolution of income inequality in Germany and Switzerland since the turn of the millennium, Markus M. Grabka and Ursina Kuhn

Lebenslagen in Deutschland Entwurf des 4. Armuts- und Reichtumsberichts der Bundesregierung, Stand 17.09.2012 17:00

85 INCOME DISTRIBUTION DATA REVIEW – GREECE

1. Available data sources used for reporting on income inequality and poverty as well as wage inequality

1.1. OECD reporting:

The OECD Income Distribution Data for Greece are computed using data from the Household Budget Survey (HBS) and are provided by the Bank of Greece.

In the OECD database, income inequality and poverty rates are currently available for income years 1974, 1988, 1994, 1999, 2004, 2008 and 2009.

1.2. National reporting and reporting in other international agencies:

 Hellenic Statistical Authority is the official national survey in Greece. The Greek Survey on Income and Living Conditions is part of the European Statistical Program and has replaced since 2003 the European Community Household Survey (ECHP). The survey is the reference for comparative statistics on income distribution and social exclusion in the European Union.

 Greece has been included in the EU-SILC (Statistics on Income and Living Conditions) survey since 2003 onwards (income year 2002). EU-SILC is a multi-dimensional instrument focused on the income and the living conditions of different types of households. It is collecting, on an annual basis, timely and comparable multidimensional micro-data on income, material deprivation, housing condition, labour, education, health and subjective well-being.

 EUROSTAT has been computing indicators on inequalities and poverty for Greece from 2000 (income year 1999) onwards.

 The Luxembourg Income Study Database (LIS) included Greece in years 1995, 2000, and 2004. It is based on the Microcensus survey that is presented in more details in the below table.

The below table presents the main characteristics of those four datasets:

86 Table 11. Characteristics of datasets used for income reporting, Greece

OECD reference series LIS database Statistics Greece Eurostat income distribution database Name Household Budget Household Income and Survey on Income and EU-SILC Survey Living Conditions Survey Living Conditions / ECHP (1995 & 2000) Survey on Income and Living Conditions / EU- SILC (2004) Name of the National level: General Hellenic Statistical Eurostat responsible agency Secretariat of the Authority National Statistical Bank of Greece Service of Greece (1995 & 2004) National Statistical Office (2004) Year (survey and 1974, 1988, 1994, 1999, 1995, 2000, 2004 1999-2010 Survey 2000 to 2011 income/wage) 2004, 2008, 2009 representing income for years 1999 to 2010 Period over which Income in the previous Annual income Annual income N-1 Annual income N-1 income is assessed year for self-employed. Month for wages, salaries and transfers. Monthly income is converted to an annual equivalent by multiplied by 12. Covered population All adults. The reference all persons living in The primary sample population is all citizens private households. units (PSUs) are the officially living on the Persons living in areas (one or more Greek territory in private collective households unified building blocks). households. Persons and in institutions are The secondary sample living in generally excluded units (SSUs) are the collective households from the target households. and in institutions are population. excluded from teh target population, as wells as households having members in diplomatic mission Sample size 2008: 3460 Households The initial sample size The initial sample size / 8930 individuals The final sample was 8.000 households was 8.000 households (sampling fraction: included 6,928 (2009) (2009) 1/1000) households, of which 5,568 completed the questionnaire (2005 survey) Sample procedure Cross-section household Two-stage design with stratified two-stage stratified two-stage surveys (from 2008 panel first stage unit sampling design sampling design and rotation sample) stratification Response rate 68,5% (2008) 80.79% 89.2% (2009) 89.2% (2009) Imputation of Within completed No imputation No imputation missing values questionnaires, all items procedure was procedure was applied. are fully imputed. applied. Unit for data Household Individual Household Household collection Break in series No No No No

Web source: http://www.oecd.org/els/ http://www.lisdatacenter. http://www.statistics.gr http://epp.eurostat.ec.eu socialpoliciesanddata/in org/wp- /portal/page/portal/ES ropa.eu/portal/page/port comedistributionandpov content/uploads/our-lis- YE/BUCKET/A0802/O al/income_social_inclusi ertydatafiguresmethodsa documentation-by-gr04- ther/A0802_SFA10_M on_living_conditions/qua ndconcepts.htm survey.pdf T_AN_00_2011_00_2 lity/national_quality_repo

011_01E_F_EN.pdf rts

87 2. Comparison of main results derived from sources used for OECD indicators (=benchmark) with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below figure shows the evolution of Gini coefficients for Greece from 1990 to 2010, as reported by the OECD, LIS, Hellenic Statistical Authority (HSA), and the EU-SILC.

Figure 16.1 Trends in Gini coefficient (disposable income)

OECD reference series (HBS) LIS (EU-SILC) HSA (EU-SILC) Eurostat (EU-SILC)

0.40

0.35

0.30 Gini (at disposable income) (at disposable Gini

0.25 1990 1995 2000 2005 2010

According to the OECD series, income inequality in Greece rose slightly during the 1990s, before dropping considerably during the 2000s, reaching 0.307 in 2008. For comparison, the latest data point for 2009 in the OECD series is shown, based on EU-SILC data.

While data from HBS suggest a continuous decline of the Gini coefficient, data from EU-SILC suggest an inverse u-shape trend with little change in the level (albeit an increase in 2009). It is difficult to depict the reasons for this divergency.

The LIS series, while only available until 2004 shows trends exhibiting a decline in inequality levels that is similar to the OECD series, for the period between the mid-1990s and mid-2000s, with 0.327 points in 2004, versus 0.307 for the OECD in 2004.

Also, when comparing the income quintile share ratio (S80/S20) from the OECD series with the series from the EU-SILC and the HSA, it is visible that the two series diverge.

88 Figure 1.2 S80/S20

OECD reference series (HBS) HSA (EU-SILC) Eurostat (EU-SILC)

7.0

6.5

6.0

5.5 S80/S20

5.0

4.5

4.0 1990 1995 2000 2005 2010

Regarding the P90/P10 Index, data is only available from the OECD and the LIS series. Both show similar levels and trends, displaying a constant decline for the LIS series since 1995, and for the OECD series since 2000, reaching 4.0 points in 2008.

Figure 1.3 P90/P10

OECD reference series (HBS) LIS (EU-SILC)

6.0

5.5

5.0

4.5 P90/P10

4.0

3.5

3.0 1990 1995 2000 2005 2010

89 2.1.2 Time series of poverty rates

According to the OECD series, poverty rates dropped since 1994 in a continuous way, from some 14% to some 11% in 2008. For comparison, the latest data point for 2009 in the OECD series is shown, based on EU-SILC data. This suggests a poverty rate of rather 13%.

While the LIS series reports an event stronger decline from 1996 to 2004, Eurostat and HAS data differ, in some years significantly..

Figure 2.1 Trends in poverty rates

OECD reference series (HBS) LIS (EU-SILC) HSA (EU-SILC) Eurostat (EU-SILC) 17%

16%

15%

14%

13%

12%

11% Poverty rate (at 50% median income) median (at 50% rate Poverty

10% 1990 1995 2000 2005 2010

As for child poverty, the OECD series remains broadly stable. For child poverty levels, the difference is very pronounced with Eurostat estimates: 12% versus 16%.

Figure 2.2 Trends in Child poverty rates

OECD reference series (HBS) LIS (EU-SILC) Eurostat (EU-SILC)

20%

19%

18%

17%

16%

15%

14%

13%

12%

11% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 10% 1990 1995 2000 2005 2010

90 2.2 Wages

See Part II of the present Quality Review

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Those are not comparable to the shares computed on the basis of EU-SILC. Indeed, HBS data report incomes net of taxes while EU-SILC reports income taxes.

Table 12. Shares of income components in total disposable income, OECD reference series

Figure 3. Trends in shares of public social transfers

91 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and the other income series:

Equivalence scale: The OECD reference series, as well as the LIS series, use the square root of household size, whereas the EU-SILC series and Hellenic Statistical Authority (HSA) series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

The different indicators for Greece are generally quite dissimilar, particularly between the OECD series and the EU-SILC and HSA series (the latter being based on EU-SILC). There is less difference between the OECD and the LIS series but the latter go up only to 2004. Overall, the EU-SILC data show consistently higher levels of poverty and inequality rates, with more fluctuating trends. Moreover, the OECD series shows a decline in inequality and poverty rates for Greece until 2008.

There is some agreement among data providers that the trend data since 2004 are more plausible depicted by EU-SILC data. As the OECD database includes 2004 data on the basis of both data sources, it may be reconsidered to adjust pre 2004 data and move the OECD reporting to EU-SILC.

92 INCOME DISTRIBUTION DATA REVIEW – HUNGARY

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD is using data from the Hungarian Household Panel and subsequently from the Household Monitor Survey provided by the Tarki Social Research Institute. In the OECD database, income inequality and poverty rates are currently available for the years 1991, 1995, 2000, 2005, 2007 and 2009.

1.2. National reporting and reporting in other international agencies:

 EUROSTAT has been computing indicators on inequalities and poverty rates for Hungary from 2000 (income year 1999) onwards.

 Hungary has been included in the EU-SILC (Statistics on Income and Living Conditions) survey since 2000 onwards (income year 1999). EU-SILC is a multi-dimensional instrument focused on the income and the living conditions of different types of households. It is collecting, on an annual basis, timely and comparable multidimensional micro-data on income, material deprivation, housing condition, labour, education, health and subjective well-being. Every year, both cross-sectional data (pertaining to a given time or a certain time period) and longitudinal data (pertaining to individual-level changes over time, observed periodically over a four year period) are collected.

 The Luxembourg Income Study Database (LIS) includes Hungary in years 1990, 1993, 1998 and 2004. It is based on the Household Monitor Survey that is presented in more details in the below table.

 Unfortunately, data on inequality and poverty indicators from the Hungarian Central Statistical Office is unavailable at this time.

The below table presents the main characteristics of those three datasets:

93 Table 13. Characteristics of datasets used for income reporting, Hungary

OECD reference series income Eurostat (EU-SILC) LIS database distribution database (TARKI)

Name Household Monitor Survey (2007 + EU-SILC Household Monitor Survey 2009) Hungarian Household Panel (1991 – 1995) Name of the Tarki Social Research Institute Eurostat Tarki Social Research Centre responsible agency

Year (survey and 1991-2009 (missing 2001, 2002, 2004, 2000-2003 and 2005-2010 1991, 1994, 1999, 2005 income/wage) 2006 and 2008) representing 1999-2002 and representing 1990, 1993, 1998 2004-2009 income and 2004 income years. Period over which For most income types respondents Annual income for the all Annual income for the all year N-1 income is assessed report the number of months they had year N-1 the specific income type during the 12 month period before the survey, and the average amount in those months. In case of more are income types (eg. bonuses, premiums at work, capital incomes) annual amount is reported. Covered population Agricultural and general government workers are EXCLUDED

Sample size Unweighted sample size 2024 For 2010 survey: Actual For 2005 survey: 2,058 households. sample size of 11500 interviewed households containing households, with an 5,284 individuals (of which 3,808 achieved sample size of adult individuals who completed 9813 households. the interview, 639 adult individuals who did not complete the interview, and 837 children under 16 years old). Sample procedure Longitudinal (HHP) and cross-sectional Stratified sampling Cross-sectional (HMS) according to different design by rotational group Response rate 44.6% (2007 income survey) 85% 72% Imputation of Missing data on specific income types Imputation of zero and missing missing values are not imputed. values of total personal current income for employed, self- employed and pensioners. Imputed values based on collected data. Unit for data Household Household Mostly at the individual level, collection except for income from household farms/enterprises, rental income, investment income, childrelated benefits, and support payments from government and other households as well as the irregular lump -sums (prizes and premiums).

Break in series Data for 1981, 1983 and 1990 are estimations by INSEE. Earnings are net of employee social security contributions but not of income tax.

Web source: http://www.oecd.org/els/socialpoliciesan http://epp.eurostat.ec.europa http://www.lisdatacenter.org/wp- ddata/incomedistributionandpovertydataf .eu/portal/page/portal/incom content/uploads/our-lis-

iguresmethodsandconcepts.htm e_social_inclusion_living_co documentation-by-hu05-survey.pdf nditions/quality/national_qua

lity_reports

94 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below figure shows the evolution of Gini coefficients for Hungary from 1990 to 2010, as reported by the OECD, the EU-SILC and LIS.

Figure 1.1 Trends in Gini coefficient (disposable income)

According to the OECD reference series, income inequality in Hungary was broadly stable over the last twenty years, with a Gini between 0.27 and 0.30.

The OECD reference series and the Eurostat series show contrasting levels and trends. The OECD reference series shows a rather steady trend throughout, whereas the Eurostat series in particular show a peak in 2005 with the Gini reaching 0.333 before falling down below levels recorded by the OECD series at 0.241 in 2009. This peak remains unexplained. However, in 2010 the Eurostat series is similar in level to the OECD reference series in both poverty and inequality indicators.

95 Figure 17.2 S80/S20

2.1.2 Time series of poverty rates

According to the OECD series, where data on poverty rates is available from 1991 to 2009, levels have remained stable from 6.3% in 1991 to 6.8% in 2009. However, levels for Child poverty rates have slightly dropped from 10.3% in 1996 to 9.4% in 2009. Again, the Eurostat series peaks remarkably in 2005 before dropping to levels close to the OECD reference series.

Figure 18.1 Trends in poverty rates

96 Figure 2.2 Trends in Child poverty rates

2.2 Wages

See Part II of the present Quality Review

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for 2008, according to the OECD benchmark series (GSOEP) and according to EU-SILC. Those are not comparable to the shares computed on the basis of EU-SILC. Indeed, HBS data report incomes net of taxes while EU-SILC reports income taxes.

Table 14. Shares of income components in total disposable income, OECD reference series

97 Figure 19. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and other sources

The OECD reference series uses the square root of household size, whereas the EU-SILC series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

Hungary is one of the few OECD countries included in the database for which income estimates only on a basis net of taxes are available. This does not impede on the indicators shown above based on disposable income but disallows some redistributive analysis.

5. Summary evaluation

The different indicators for Hungary based on HSM and on EU-SILC are generally quite dissimilar, in particular in one year (2005). This discrepancy cannot be accounted for.

98 INCOME DISTRIBUTION DATA REVIEW - IRELAND

1. Available data sources used for reporting on income inequality and poverty

1.1 OECD Reportings

The OECD have been using two types of data sources for income reporting on Ireland. First, the OECD was relying on the Living in Ireland Survey (LIIS) from the Central Statistics Office which data have been included in the OECD Database for the years 1987, 1994 and 2000. From 2004, the OECD is using data from the EU Survey of Income and Living Conditions (EU-SILC) for 2004, 2006, 2007, 2008 and 2009. The change in survey in 2004 was concerning data from 2003. It corresponds to the change of survey within the Central Statistics Office in Ireland from LIIS to EU-SILC.

However, OECD considers that there is a strict break in series due to the change in source to EU- SILC in 2004. Data prior to 2004 are strictly not comparable with data from the 2004 onwards.

1.2 National Reporting and Reporting in other international agencies

1.2.1 National reporting:

Income distribution and poverty indicators for Ireland at the national level are only available from this survey:  Up to 2003 - All the indicators concerning income inequalities and poverty rates are gathered in the Living in Ireland Survey up to 2003. The Living in Ireland Survey is the Irish component of the European Household Panel Survey (ECHP) which is coordinated by Eurostat. The Economic and Social Research Institute (ESRI) was responsible for this survey. Some more details on income collection concerning the Living in Ireland Survey: incomes collected are mostly individual, excl. housing allowances, social assistance, rental income and inheritance/lottery winning. Most variables are collected net of taxes and contributions (with the exception of self- employment earnings, and wages which are collected also gross).

1.2.2 International reporting:

In addition to the OECD time-series, the following reportings are available on income inequalities and poverty for Ireland at an international level:

 From 2003 onwards - The Survey on Income and Living Conditions (EU-SILC) of Eurostat is presenting these results. The production of National Statistics for this survey is handled by the Central Statistics Office. Data are available annually from 2003 up to 2010.

 Eurostat is also publishing indicators on income inequalities and poverty for Ireland from 1995 onwards, based on the ECHP (itself based on LIIS).

 Ireland is also included in the Luxembourg Income Study Database (LIS). LIS is using the Survey of Income Distribution, Poverty and Usage of State Services in 1987, the Living in Ireland Survey in 1994, 1995, 1996 and 2000. For 2004 (last available year on LIS), the Survey on Income and Living Conditions / EU-SILC has been used.

99 Table 15. Characteristics of dataset, Ireland

Name OECD reference series since Living in Ireland Survey Luxembourg Income Study Database 2004 (EU-SILC ) (LIIS) (LIS) Name of the Eurostat / Central Statistics Office Economic and Social Research LIS responsible (CSO) Institute (ESRI) / European Household Panel Survey (ECHP) Goal Official source of data on To provide an up-to-date and To enable, facilitate, promote, and incomes and living conditions of comparable data source on conduct cross-national comparative different types of households. The personal incomes. research on socio-economic outcomes survey also collects information and on the institutional factors that on poverty and social exclusion. shape those outcomes. Year From 2003 onwards Up to 2003 1987 - Survey of Income Distribution, Since 2004, this survey is Wave 1: 1994 Poverty and Usage of State Services conducted throughout the Wave 8: 2001 (last one) 1994, 1995, 1996 and 2000 - the Living European Union in Ireland Survey 2004 – EU SILC Data Annually Annually (from June to Refer to the survey of references collecting December) according of the concerned year frequency Covered Private households - The All private households in the Refer to the survey of references population sampling frame for the SILC national territory (incl. according of the concerned year survey was drawn from the 2006 collective households but excl. Census of Population institutional ones) Sample size 12,641 individuals in 2009 4,048 households and 14,589 Refer to the survey of references 11,587 individuals individuals in 1994 according of the concerned year Minimum sample size: 3750 2,865 households and 9,131 households for cross-sectional / individuals in 2001 2750 for longitudinal; 8000 individuals for cross-sectional / 6000 for longitudinal Sampling A representative random sample Two-stage sampling with Refer to the survey of references method of households throughout the District Electoral Divisions according of the concerned year country is approached to provide (DED) selected systematically the required information. The within each stratum and survey is voluntary from a households of the electors respondents perspective; nobody selected within each DED. can be compelled to co-operate. Disseminati 12 months after the end of the Refer to the survey of references on reference year according of the concerned year frequency Sampling Individuals and Households Individuals and Households Refer to the survey of references unit according of the concerned year Response Unknown Between 57% (1994) and 88 Refer to the survey of references rates (1997) according of the concerned year Remark The first results for 2003 In 2000, the Irish sample of Refer to the survey of references (published in January 2005) have individuals and households according of the concerned year been revised following the followed from Wave 1 was application of improved re- supplemented by the addition of weighting and calibration 1,500 new households to the methods that are in line with EU total. These additional recommendations. The effect of households, as well as the the revisions has been to lower original sample, were followed both the risk of poverty and in 2001. consistent poverty measures. Websource http://www.statcentral.ie/viewStat http://www.esri.ie/ http://www.lisdatacenter.org/our- .asp?id=137 http://www.esri.ie/pdf/WP016 data/lis-database/by- http://www.cso.ie/en/index.html 3%20Living%20in%20Irelan country/ireland-2/ d%20Survey.pdf http://www.lisproject.org/tech doc/ie/ie00survey.pdf

100

2. Comparison of main results derived from sources used from OECD indicators with alternatives sources

2.1. Income

2.1.1 Time series of Gini coefficients and other inequality indicators

It should be noted that the OECD considers the time series pre- and post EU-SILC for Ireland not comparable. In the different OECD publications (Growing Unequal, Displaced worker Survey, Divided We Stand), the trends are shown from 1987 through 2000, and from 2004 through 2008. 2004 is considered as “break” year. This Data Review is adopting the same approach.

The different series on Gini coefficients for disposal income in Ireland are broadly similar and are showing similar trends. According to the OECD series, Gini coefficients decreased between 1987 and 2000. They also decreased between 2004 and 2008, before a large surge in 2009.

Up to 2004, OECD data can be compared with Eurostat, and with the Luxembourg Income Suvey. The Luxembourg Income Suvey was presenting very similar results than the OECD from 1987 to 2004. In addition, even if very few points of reference were found in the Living in Ireland Survey, they show levels in the order of magnitude of the OECD data. Please note that the Living in Ireland Suvey was compiling two different Gini coefficients in 2003 as the methodology of calculation had changed to comply with EU regulation. Concerning Eurostat, its line tends to amplify the moves on Gini coefficients compared to other survey data. In 1997, income inequalities reached a peak with a Gini coefficient of 0.34 before recording a sharp decline down to 0.29 in 2000. Such a big variation is not mentioned with OECD figures which are recording a regular decline from 1994 to 2000, reaching a level at 0.304 in 2000.

From 2004 onwards, the OECD Gini coefficients can be compared with Eurostat. Eurostat series confirme the declining trend as Gini coefficients went down from 0.32 in 2003 to 0.29 in 2008. There were minor variations on these coefficients between the OECD and EU-SILC /Eurostat survey over this period. From 2008 to 2009, the two datasets are recording a sharp increase of inequalities up to 0.331.

Figure 20. Trends in Gini coefficients, Ireland (1987 – 2010)

OECD reference series (LIIS and EU-SILC) Eurostat (ECHP and EU-SILC) LIIS LIS 0.40

0.35

0.30

0.25

101 The series on the S80/S20 income share ratio are following the same trend than the Gini coefficients after 2004. However, for the years1994 to 2000, they suggest stability, rather than a decrease.

Please, note that as mentioned previously for the Gini coefficients, the Living in Ireland Suvey is compiling two slightly different income inequality ratios in 2003 as the methodology of calculation has changed to comply with EU regulation.

Figure 2. Trends in S80/S20, Ireland (1987 – 2010)

OECD reference series (LIIS and EU-SILC) LIIS Eurostat (ECHP and EU-SILC) 6.00

5.50

5.00

4.50

4.00

3.50

3.00

From 2004 onwards, the OECD time series are very similar with the ones provided by Eurostat. According to Eurstat series, this ratio remained broadly unchanged before that date, from 1994 to 2004. From 2004 onwards, there was a decreasing trend in the quintile share ratio from 5.0 in 2003/2004 down to 4.3 in 2008. The latest results published by EU-SILC (2010 year survey) showed a significant increase of this ratio: the average income of those in the highest income quintile was 5.5 times that of those in the lowest quintile in 2009. The same ratio was 4.3 one year earlier thus signifying greater inequality in the income distribution in 2009. The results calculated by Eurostat are recording the same trend: the ratio increased from 4.2 in 2008 up to 5.3 in 2009. Finally, for 2009, the OECD time series also pointed out an increase of the S80/S20 ratio up to 5.358.

From 2008 to 2009, the two datasets are recording a sharp increase of inequalities, according to both inequality indicators. While national and international observers agree on the plausibility of an increase in inequality, there have been some doubts expressed as to the magnitude of this increase (plus 4.4 points in Gini, plus 1 point in S80/S20 share ratio). One issue may be concern that the sample for the 2009 income data over-represents the better-educated compared with 2008.

2.1.2 Time series of poverty rates, poverty composition

The analysis of the OECD time series on poverty rates is showing an increase in poverty rates from 1987 to 2000 (especially between 1994 and 2000), where a peak is recorded for all databases, and a regular decrease between 2004 and 2009. This can be confirmed by analyzing poverty rates with a 50% threshold of median income and with a 60% threshold of median income.

102 According to the OECD reference survey, the share of the Irish population living with less than 50% of the median equivalised income increased from 10.6% in 1987 to 15,4 % in 2000. This rate is estimated at 9% in 2009. The Luxembourg Income Survey is presenting very similar rates and shares the same patterns up to 2004 (latest year in LIS database).

The Eurostat figures are also showing a similar trend. The share of the Irish population living with less than 50% of the median equivalised income increased from 7% in 1994 to 15% in 2000. Note that for year 1994, the poverty rate as recorded by Eurostat was significantly lower than the OECD one with poverty rates, respectively estimated at 7% (Eurostat) and at 11% (OECD). After the 2000’s peak when OECD and Eurostat were recording the same levels, Eurostat had constantly recorded lower poverty rates than the OECD. There was a difference of about 2 points in the years 2005 to 2008.

From 2008 to 2009, Eurostat reported that there was a slight increase of poverty rates (from 7.30% to 7.80%) whereas the OECD recorded a slight decrease on the same indicato (from 9.10% to 8.99%). This divergence of results led to a reduction of spread between Eurostat and OECD on this indicator.

Figure 21. Trends in Poverty rates, 50% threshold, after taxes and transfers, Ireland (1987 – 2010)

OECD reference series (LIIS and EU-SILC) Eurostat (ECHP and EU-SILC)

LIS - Relative Poverty Rates - Total Population (50%)

18% 16% 14% 12% 10% 8% 6% 4% 2%

1986 1987 1990 1991 1992 1995 1996 1997 1999 2000 2001 2004 2005 2007 2008 2009 1988 1989 1993 1994 1998 2002 2003 2006 2010 2003 revised 2003

The analysis of poverty rates with a 60% threshold is presenting similar features: an increase from 1987 to 2000, then a regular decline from 2004 up to 2008. However, figures from 2008 to 2009 are showing an increase in the rates of poverty according to Eurostat whereas the OECD is confirming a declining trend.

From 1987 to 2000, and according to the OECD time series, the share of the Irish population living with less than 60% of the median equivalised income increased from 19.51% in 1987 to 23.30% in 2000. These rates were very similar to the Luxembourg Income Survey ones which recorded a rise from 19.99% to 22.49%. In 2000, the Living in Ireland Survey calculated a poverty rate estimated at 20.90% which was relative close to the OECD rates. Eurostat also compiled similar poverty rates. According to Eurostat dataset, poverty rates with a 60% threshold remained broadly unchanged from 1994 to 1999 with a 19% rate before recording an increase of 2% points in 2000.

Please, note from 1995 to 2009, the OECD rates were higher than the ones published by Eurostat. Indeed, from 1995 to 2009, Eurostat published constantly lower poverty rates at a 60% poverty line (with

103 an approximate spread of 1.5 percentage points) compared to the OECD ones. The recent 2009 results enabled to reduce this spread between OECD and Eurostat/EU-SILC.

According to Eurostat, the share of the Irish population living with less than 60% of the median equivalised income increased from 2008 to 2009 to reach 16.10%. The 2009 increase did not show up in the OECD time-series which recorded a decrease between 2008 and 2009 (16.85% to 16.18%), following the declining trend from 2000.

Please note that the Living in Ireland Survey in 2000 published several figures on poverty rates (ie, Percentage of persons below median relative income poverty lines (based on income averaged across individuals) according to different equivalence scales.

- Equivalence Scale A: 1 for the first adult, 0.66 for each other adult and 0.33 for each child; - Equivalence Scale B: 1 for the first adult, 0.6 for each other adult and 0.4 for each child; - Equivalence Scale C : 1 for the first adult, 0.7 for each other adult and 0.5 for each child. The official figures which are published for Ireland and are also shown in this report are based on Scale A. This scale is slightly different from the one for the OECD and may help explain the fact that inequalities are larger according to the OECD time series.

Figure 22. Trends in Poverty rates, 60% threshold, after taxes and transfers, Ireland (1987 – 2010)

OECD reference series (LIIS and EU-SILC) Eurostat (ECHP and EU-SILC)

LIIS - 60% median income threshold LIS - Relative Poverty Rates - Total Population (60%)

24%

22%

20%

18%

16%

14%

12%

10%

1987 1989 1990 1992 1993 1995 1997 1998 2000 2001 2003 2004 2005 2007 2008 2010 1986 1988 1991 1994 1996 1999 2002 2006 2009

2003 revised 2003

Children are the most vulnerable and affected by poverty in Ireland. According to the EU-SILC 2010 results (for the income year 2009), and using a threshold of 60% of the median, almost one in five children were at risk of poverty in 2009 compared with almost one in ten of the elderly population.

104 Figure 23. Trends in child poverty rates, after taxes and transfers, Ireland (1987 – 2010)

OECD - Children Poverty Rate - Age group 0-17 Eurostat - Children Poverty Rate - Age group 0-18 _ Pov Line 50% Eurostat - Children Poverty Rate - Age group 0-18 _ Pov Line 60% 25%

20%

15%

10%

1988 1990 1991 1993 1996 1998 1999 2001 2003 2005 2006 2008 2010 1986 1987 1989 1992 1994 1995 1997 2000 2002 2004 2007 2009

2003 revised 2003 The OECD figures on poverty rate are calculated with a 50% median income threshold and were very similar to the Eurostat ones since 2004. However, for the year 2008, the OECD and Eurostat presented more diverging results with a difference of 2.8 percentage points: 8.6% for Eurostat versus 11.4% for the OECD. In 2009, OECD recorded a larger decrease than Eurostat : children poverty rate was estimated at 10.2% for the OECD versus 8.50% for Eurostat. Please, note that for the period before and after 2004, the OECD considers figures as not comparable.

2.2. Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference survey 2009 natcur 12600 6335 3 18937 468 2875 9336 -5120 26496 % av HDI 47.6% 71.5% 1.8% 10.8% 35.2% -19.3% Figure 6 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show pretty similar trends throughout the period.

105 Figure 6 Trends in shares of public social transfers

40 35 30 25 20 15 10 5 % transfers in disposable income (OECD reference… 0 1990 1995 2000 2005 2010

4. Metadata of data sources which could explain differences and inconsistencies

OECD reference series match with the other data sources available for income inequalities and poverty rates with a few exceptions:

- For Gini coefficients, Eurostat tends to amplify the move with a larger sensibility both to upward and downward trends than the OECD time-series;

- For poverty rates to 50% and to 60% income thresholds, Eurostat/EU-SILC recorded lower levels of poverty rates than the OECD ones for all the period under consideration. On one hand, all the poverty rates with a poverty line of 50% reached the same levels in 2000 while being slightly different before and after (OECD compiling higher levels than Eurostat/EU-SILC). On the other hand, considering a poverty line of 60%, the OECD recorded higher rates than Eurostat/EU-SILC over all the period under consideration.

- The Eurostat/EU-SILC results presented an increase of poverty rates from 2008 to 2009, reversing the previous declining trend, whereas the OECD continued to record a slight decrease of poverty rates in 2009.

5. Summary evaluation

There is a nearly perfect accordance between the OECD time-series and the other data sources for income inequalities (Gini and S80/S20). Somewhat more divergences can be recorded concerning poverty rates. However, the last results from 2009 led to a reduction of spread between different datasets.

There remain two issues for further clarification:

- Recent results of all databases for the income year 2009 suggest a very sharp increase of income inequality. Careufl verification is needed as to the magnitude of this increase (plus 4.4 points in Gini, plus 1 point in S80/S20 share ratio, in one year). There might be an issue of change in the sample design (see above).

- Second, there may be an issue with the estimates for pre-tax and transfers incomes. While, according to the LIIS database (2000), the difference between pre- and post tax/transfer inequality matched the OECD average, the results based on EU-SILC (2004, 2008) suggest a huge redistributive effect of taxes and transfers, higher than in any other OECD country (table 3). As

106 such changes are very significant and not being confirmed by national studies, OECD is, for the time being, not publishing the pre-tax/transfer inequality estimates. This issue, nevertheless, merits rapid clarification.

Table 3. Estimates of pre- and post-tax/transfer Gini coefficients, total population

2000 2004 2008 Gini of disposable income 0.304 0.314 0.293 Gini before taxes and transfers 0.434 0.504 0.537 Redistribution (%) 30% 38% 45%

107 INCOME DISTRIBUTION DATA REVIEW – ICELAND

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Iceland are computed from EU-SILC data from 2004 onwards. The data are computed internally and sent to Statistics Iceland for verification. In the OECD database, income inequality and poverty rates are currently available for income years 2004 to 2009.

1.2. National reporting and reporting in other international agencies:

 EUROSTAT has been computing indicators on inequalities and poverty rates for Iceland from 2004 (income year 2003) onwards.

 Iceland has been included in the EU-SILC (Statistics on Income and Living Conditions) survey since 2003 onwards (income year 2002). EU-SILC is a multi-dimensional instrument focused on the income and the living conditions of different types of households. It is collecting, on an annual basis, timely and comparable multidimensional micro-data on income, material deprivation, housing condition, labour, education, health and subjective well-being. Every year, both cross-sectional data (pertaining to a given time or a certain time period) and longitudinal data (pertaining to individual-level changes over time, observed periodically over a four year period) are collected.

 Statistics Iceland is the centre for official statistics in Iceland and collects, processes and disseminates data on the economy and society. It uses data provided by EU-SILC.

The below table presents the main characteristics of those three datasets:

108 Table 16. Characteristics of datasets used for income reporting, Iceland

OECD reference series income Eurostat Statistics Iceland distribution database Name EU-SILC EU-SILC EU-SILC Name of the Eurostat Statistics Iceland responsible agency Eurostat

Year (survey and 2004-2009 2004-2001 survey representing 2004-2001 survey representing income/wage) income for years 2003-2010 income for years 2003-2010

Period over which Annual income in the previous Annual income N-1 Annual income N-1 income is assessed year, also in case of transfers from public sources Covered population The Population register The Population register The Population register Sample size 3021 households (2010) 3021 households (2010) 3021 households (2010)

Sample procedure cross-section Simple random sampling Simple random sampling Response rate 78.00% 78.00% 78.00% Imputation of missing No missing values, negative values values treated as suggested in the terms of references Unit for data collection Household Individual Individual Break in series no no no

Web source: http://www.oecd.org/els/socialpol http://epp.eurostat.ec.europa.eu/p http://www.statice.is/Statistics/Wa iciesanddata/incomedistributiona ortal/page/portal/income_social_in ges,-income-and-labour-

ndpovertydatafiguresmethodsan clusion_living_conditions/quality/n market/Income-distribution

dconcepts.htm ational_quality_reports

109 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below figure shows the evolution of Gini coefficients for Iceland from 1990 to 2010, as reported by the OECD, the EU-SILC and Statistics Iceland.

Figure 24.1 Trends in Gini coefficient (disposable income)

OECD reference series (EU-SILC) Statistics Iceland (EU-SILC) Eurostat (EU-SILC) 0.36

0.34

0.32

0.30

0.28

0.26

0.24 Gini (at disposable income) (at disposable Gini

0.22

0.20 1990 1995 2000 2005 2010

According to the OECD reference series, income inequality in Iceland rose over the last decade, reaching 0.301 points in 2008, before a sudden decline in 2009 to 0.266 points.

The EU-SILC and Statistics Iceland series are identical since Statistics Iceland uses the methodology and data of EU-SILC. The levels and trends of the latter series are very similar to the OECD series, exhibiting slightly lower levels throughout. Contrary to the OECD series, the EU-SILC and Statistics Iceland series show Gini levels for year 2010, which exhibits a remarkable decline from 0.296 in 2008 to 0.236 in 2010.

Also, when comparing the income quintile share ratio (S80/S20) from the OECD series with the series from the EU-SILC and Statistics Norway, the trends are overall quite similar with the OECD showing higher levels throughout. We can observe a general rise in S80/S20 levels from 2003 to 2008, with the OECD series reaching 4.4 points in 2008.

110 Figure 1.2 S80/S20

2.1.2 Time series of poverty rates

According to the OECD series, where data on poverty rates is available from 2004 to 2009, levels have risen from 5.9% in 2004 to 6.3% in 2009. There are, however, inconsistencies between OECD and the alternative series after the year 2006 which needs to be investigated further: in 2007, the OECD poverty rate increased while the Eurostat series inclined and in 2008 and 2009, the inverse happened.

The series for EU-SILC and Statistics Iceland show fluctuating levels of poverty rates since 2004. Indeed, poverty rates in Iceland dropped 1 percentage point between 2004 and 2008, reaching 4.5%, before rising again to reach 5.4% in 2010. Also, the levels for EU-SILC and Statistics Iceland are consistently lower than the OECD series.

Figure 2.1 Trends in poverty rates

OECD reference series (EU-SILC) Statistics Iceland (EU-SILC) Eurostat (EU-SILC) 10%

4% Poverty rate (at 50% median income) median (at 50% rate Poverty

2% 1990 1995 2000 2005 2010

111 As for child poverty, data is available from the OECD and EU-SILC from 2004 to 2009. Moreover, the EU-SILC series exhibits substantial variations in levels and trends, with a decline from 6.5% to 4.1% between 2004 and 2008, before rising again and reaching a peak at 6.5% in 2010. This being said, the trends of the two series appear consistent except for the year 2008, both indicating a sharp rise after 2008. Again, the levels for EU-SILC are consistently lower than the OECD series.

Figure 2.2 Trends in Child poverty rates

OECD reference series (EU-SILC) Eurostat (EU-SILC)

10%

3% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 2% 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Unfortunately, such information is not available for the other data sources.

Table 17. Shares of income components in total disposable income, OECD reference series

Self Survey Year Unit Wages Capital Employment Transfers Taxes Disposable income EU-SILC (OECD-ELS) 2008 natcur 28,529 5,473 1,002 4,435 -10,618 4,189,193 % av HDI 98% 19% 3% 15% -36%

112 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and other sources

The OECD reference series uses the square root of household size, whereas the EU-SILC series and Statistics Iceland series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

The different Gini series for Iceland are generally very similar, even though it would be interesting to have data for the OECD series after 2009 in order to confirm the continuous substantial decline as observed in the EU-SILC and Statistics Iceland series. Concerning poverty rates, data for the OECD series is limited and it is thus difficult to objectively compare. Nonetheless, for the years 2004 to 2009, the levels of the OECD and EU-SILC series are quite dissimilar, as the OECD series is consistently higher than the EU-SILC series. Moreover, there are inconsistencies between the OECD and the Eurostat series for poverty rates.

113 INCOME DISTRIBUTION DATA REVIEW – ISRAEL7

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD Income Distribution Data for Israel are provided by the Central Bureau of Statistics (C.B.S.) and based on the cross sectional household survey conducted by the Consumption and Finance division of the C.B.S. In the OECD database, income inequality and poverty rates are currently available for years 1985, 1990, 1995, 2000, 2005, 2008, 2009 and 2010.

1.2. National reporting and reporting in other international agencies:

Income distribution and poverty indicators for Israel are also available from the Luxembourg Income Study, Statistics Israel, and the National Insurance Institute.

1.2.1 National reporting:

 Statistics Israel is the official national survey in Israel and has been computing data annually since 1974 through the Central Bureau of Statistics (CBS), which is an autonomous unit within the Prime Minister’s Office. From the year 1997 the data is based on the combined survey (income + expenditure). And from the year 1997 the Gini was calculated on equivalised disposable household income.

 The National Insurance Institute of Israel (NII) issued a report in 2010 on poverty and social gaps in Israel, with data on income inequality available since 1999.

1.2.2 International reporting:

 The Luxembourg Income Study (LIS) included Israel in years 1986, 1992, 1997, 2001, 2005 and 2007. It is based on the Household Expenditure Survey put together by the Central Bureau of Statistics (CBS).

Table 1 presents the main characteristics of the different sources:

7 The statistical data for Israel are supplied by and under the responsibility of the relevant Israeli authorities. The use of such data by the OECD is without prejudice to the status of the Golan Heights, East Jerusalem and Israeli settlements in the West Bank under the terms of international law.

114 Table 18. Characteristics of dataset used for income reporting, Israel

OECD reference series LIS database Statistics Israel National Insurance income distribution Institute (NII) database Name Household Expenditure

Survey Income survey Name of the responsible agency Central Bureau of Central Bureau of Statistics Central Bureau of Statistics Statistics

Year (survey and 1985, 1990, 1995, 1986, 1992, 1997, 2001, 1990 - 2009 1999, 2002 – 2010 income/wage) 2000, 2005, 2008, 2005, 2007 2009, 2010 Period over which Yearly Monthly income is assessed Income in the previous month. Same reporting period for all income types. All income data are standardized to price level.

Covered population The survey population includes the entire urban and non-urban population (including the population of East Jerusalem) except for kibbutzim, collective moshavim and Bedouins living outside of localities. Sample size 6,173 households containing 14167 households 20,364 individuals who 15171 households (2010) (2008) completed the interview (2007) Sample procedure Cross-section Cross-sectional, with Cross-sectional household survey monthly rotation of households. Response rate 89.8% (2007) 85.5% in 2008 84.1% in 2010

Imputation of No imputation was made for missing values income that originates in the use of one’s dwelling and for various types of in-kind income (non- financial income). Unit for data Household Household Household collection Break in series No No No

Web source: http://www.oecd.org/ http://www.lisdatacenter.or CF time series http://www.btl.gov.il/ els/socialpoliciesandd g/wp-content/uploads/our- English%20Homepage ata/incomedistributio lis-documentation-by-il07- /Publications/Poverty nandpovertydatafigur survey.pdf _Report/Pages/defaul

esmethodsandconcept t.aspx s.htm

115 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below figure shows the evolution of Gini coefficients for Israel from 1990 to 2010, as reported by the OECD, LIS, NII and Statistics Israel.

According to the OECD series, income inequality in Israel rose significantly between 1990 to 2010 from 0.338 to 0.376. Overall, the other series show a similar upward trend, although the LIS series shows lower levels of income inequality, while Statistics Israel and the NII show higher levels of income inequality throughout the entire period. The NII also witnessed a decline in income inequality for 2010, while the OECD data suggest an increase.

Figure 25. Trends in Gini coefficient (disposable income)

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Israeli population living with less than 50% of the median equivalised income (36.675 shekel per year in 2008) has increased from 14.6% in 1990 to 20.9% in 2010.

The LIS and Statistics Israel series both show similar upward levels in poverty rates, with the LIS series reaching 19.4% in 2005 and the Statistics Israel series reaching 19.9% in 2008.

As for child poverty, data is only available for the OECD and LIS series. Both series show generally similar levels of increasing child poverty rates. Levels nearly doubled for the OECD series, with an increase from 14.6% in 1990 to 28.5% in 2010.

116 Figure 26.1 Trends in poverty rates

Figure 2.2 Trends in child poverty rates

2.2 Wages

See Part II of the present Quality Review.

117 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Unfortunately, such information is not available for alternative data sources.

Table 19. Shares of income components in total disposable income, OECD reference series

Average income K SE TR TA HDI Self Disposable Survey Year Unit Wages Capital Employment Transfers Taxes income (HDI) OECD reference survey 2008 natcur 62,426 8,197 9,962 8,912 -16,146 73,350 % av HDI 85% 11% 14% 12% -22%

Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Differences in equivalence scales

The equivalence scale used by Statistics Israel is not the same as the OECD methodology (square root of household size); it is by the Israeli scale as follows:

118 Marginal Number of Number of persons in household weight per standard person persons

1 1.25 1.25 2 0.75 2 3 0.65 2.65 4 0.55 3.2 5 0.55 3.75 6 0.5 4.25 7 0.5 4.75 8 0.45 5.2 9+ 0.40(1) (1) For each additional person.

5. Summary evaluation

Between the different series available for inequality and poverty rates for Israel, the trends are generally quite similar overall, although the national series (Statistics Israel and NII) exhibit higher levels of Gini coefficient throughout the entire period, and the LIS series suggest somewhat lower ones.

119 INCOME DISTRIBUTION DATA REVIEW – ITALY

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD data on income inequality and poverty data for Italy refer to 1984, 1991, 1993, 1995, 2000, 2004, 2006, 2007, 2008 and 2009. These estimates are based on different data sources.

 From 1984 to 1993, OECD estimates were based on data the micro-simulation model ITAXMOD95, based on the Survey on Household Income and Wealth (SHIW), conducted by the Bank of Italy.

 From 1995 to 2004, OECD estimates were based the micro-simulation model MASTRICT, maintained at ISTAT, and also based on the data from the Survey on Household Income and Wealth (SHIW). Both micro-simulation models integrated the original survey records with imputations on taxes paid by households (information on which was not available in SHIW) but, in the process, also altered some of the values on market income. Data from this model were also provided for the income year 2008.

 From 2006 onwards (survey year: 2007 onwards), OECD estimates have been based on the EU Survey on Income and Living Conditions (EU-SILC).

Therefore, several breaks need to be taken into consideration when reporting OECD time-series data. First, a break in 1995 (due to the change in the micro-simulation model). Second, a break in 2006 (due to the switch to EU-SILC survey). These two breaks are highlighted in the different graphs presented below. The OECD Secretariat has addressed these breaks through multiplicative adjustments in 2006; while no adjustment is made for 1993-1995, on account of the fact that estimates based on the two micro-simulation models for adjacent years are very close to each other.

Please note that in the OECD database, several data may be provided for the same year. The following ‘flags’ are included in the database:

- Data labelled as “old” are those based on the ITAXMOD95 micro-simulation model;

- Data labelled as “previous” are those based on the MASTRICT micro-simulation model;

- Data labelled as “current” refer to EU-SILC data from 2006 onwards, while data for previous years are based on backward interpolation from the current EU-SILC series in order to have a consistent long time-series of data. Backward interpolation was based on the MASTRICT micro- simulation model for the years 2006 to 1995; and on the ITAXMOD95 micro-simulation model for years from 1993 to 1984.

Evidence in this Review is based on the OECD “Current” data for better consistency over time.

1.2. National reporting and reporting in other international agencies:

The following datasets are available for Italy:

120  The Bank of Italy’s Survey on Household Income and Wealth provides the longest record of household income data for Italy. Data from the most recent wave (2010) are available at: http://www.bancaditalia.it/statistiche/indcamp/bilfait/boll_stat/en_suppl_06_12n.pdf. The survey has undergone a number of methodological changes, the most significant among them being the integration of income from financial assets and family cash benefits (assegni familiari) since 2005.

 ISTAT started to collect information on the distribution of household income through the European Community Household Panel (ECHP), from 1994 to 2001, and though the EU Survey of Income and Living Conditions since 2006 (survey year: 2007). The information collected through EU-SILC has traditionally referred to net income, i.e. income after taxes. Starting in 2007, ISTAT started releasing information on ‘gross’ income, which takes into account taxes and social security contributions paid by workers. Estimates of taxes paid (and gross income) are based on the microsimulation model SM2 of the University of Siena. A description of the methodological features of the imputation is available at: http://www3.istat.it/dati/catalogo/20110726_00/ Relative to SHIW, EU-SILC has a larger sample size and response rates. Estimates of income inequality based on EU-SILC display greater inequality than those based on SHIW, reflecting a higher income share of richer households and a lower share of poorer ones: the Gini coefficient for equivalised income in EU-SILC is reported by ISTAT studies at 0.396 in 2002, as compared to 37.3 in SHIW.

 National reporting on the distribution of household economic resources in Italy has typically been based on measures of household consumption, rather than income, based on data collected by ISTAT in its household budget survey. Based on this source, ISTAT calculates relative and absolute poverty rates, with and data have been available on ISTAT statistics database since 1997. The estimation of the relative poverty, is based on a poverty line (International standard of poverty Line-ISPL) defining as poor a household of two components with a consumption expenditure lower or equal to the mean per-capita consumption expenditure; for household of different sizes, an equivalence scale is used to take into account different needs. The figures on relative poverty rates provided by ISTAT are labelled as “household relative poverty incidence in percentage”. This rate was estimated at 11.1% for 2011.

At the international level, the Luxembourg Income Survey is based on the Survey of Household Income and Wealth (SHIW) from Bank of Italy. The LIS database includes micro data for the years 1986, 1987, 1989, 1991, 1993, 1995, 1998, 2000 and 2004. Conversely, Eurostat has been reporting annual indicators on income inequalities and poverty for Italy since 1995 (income year: 1994) onwards. Two specificities in terms of definition and covered population are specified in the EU-SILC methodology (2009 COMPARATIVE EU FINAL QUALITY REPORT - 2011):

- The definition of private household used in Italy refers to “Cohabitants related through marriage, kinship, affinity, patronage and affection constitute the private household”.

- Live-in domestic personnel are not included as household members in Italy. Concerning these persons, only some socio-demographic information is collected (date of birth, sex, marital status, and duration of stay in the household).

The below table presents the main characteristics of those datasets:

121 Table 20. Characteristics of dataset, Italy

Name Survey on Household Luxembourg Households Budget Income and Wealth EU-SILC Income Study Survey (SHIW) Name of the Banca d’Italia Eurostat ISTAT Banca d’Italia responsible (based on the agency Household Income and Wealth Survey - Indagine sui Bilanci delle Famiglie Italiane) Year Every two year. Annually since 2004 Annually since 1997 (it Usually every two Microdata have been (income year : 2003) was existing since 1953 years but with some available since 1977 but has been completely irregularities since Last wave available: 2010 renewed in 1997) 1986 (1986, 1987, 1989, 1991, 1993, 1995, 1998, 2000 and 2004) Data Between January and Between May and Household Budget February to July collecting September December Survey is collecting data quarterly Covered All Italian households All Italian households All households in population excluding people living in registered in the the national barracks, rest homes and municipalities. territory, with the hospitals (estimated at 7 NB: Live-in domestic exception of per thousand of the total personal (au pairs) are not Institutions and resident population – included as household people not 2008 wave) members in Italy. registered on Concerning these persons, municipal registers only some socio- (e.g illegal demographic information immigrants)- Wave is collected (date of birth, 2004 sex, marital status, and duration of stay in the Before 2004, all household). The number private households of these persons are included in the included in the sample covered population was 62 (0.13% of interviewed individuals – wave 2010).

Sample size Approx. 8000 households The minimum sample size Approx. 24 000 Same than the with 24 000 individuals is 7250 households (15 households Survey Household 7,951 households (2010); 500 individuals aged 16 Income and Wealth 7,977 (2008); 7,768 or over). (SHIW) (2006); 8012 (2004); Achieved sample size 8012 households 8011 (2002); 7147 (2010): 19 147 (20 581 individuals (1998); 8135 (1995); – wave 2004). 8089 (1993); 8188 8001 households (1991); 8274 (1989); (19 209 individuals 8027 (1987). – wave 2000). Method of Mainly by Computer- From registers and Same than the data Assisted Personal interviews (for interviews, Survey Household collection Interviewing program single mode of data Income and Wealth (CAPI – 84% of the collection: Paper-Assisted (SHIW) interviews in 2010). The Personal Wave 2004 -73% remaining interviews are Interview (PAPI)) with CAPI and the conducted using paper- remaining using based questionnaires paper-based (PAPI, Paper-And-pencil questionnaires.

122 Personal Interviewing) Sampling Two stage stratified: Stratified two-stage Same than the method municipalities and then clustered sampling Survey Household households design, with systematic Income and Wealth sampling of households (SHIW) from municipalities Sampling Households and Municipalities (primary Households Same than the unit Individuals sample unit) and Survey Household households (secondary Income and Wealth sample unit) (SHIW)

Imputation Yes (less than 4% in n/a Yes. of data 2010) Between 5% and 10% for the most complex questions (wave 2000) Response 52.7% (2010 wave) 80.25% (Household n/a rates response rate – Total sample - 2010 wave)

Remark n/a Annual income of the The Household Budget n/a year prior to the survey Survey is the input Before 2004, Eurostat was source for the relative compiling statistics for poverty and social Italy since 1994. exclusion indicators analysis. Websource http://www.bancaditalia.it http://epp.eurostat.ec.euro http://dati.istat.it/?lang= http://www.lisdatac /statistiche/indcamp/bilfai pa.eu/portal/page/portal/in en and enter.org/our- t/boll_stat come_social_inclusion_li http://siqual.istat.it/SIQ data/lis- ving_conditions/introducti ual/unita.do?id=888891 database/by- on 6 country/italy-2/

2. Comparison of main results from OECD and alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Time series on Gini coefficients are estimated by various institutions which enable comparisons over the long time period comprised from 1984 to 2010. The Figure below shows time-series of the Gini coefficients from:

- the OECD database, with breaks shown for 1995 and 2006

- the Luxembourg Income Survey

- the Bank of Italy (with the Survey on Households Income and Wealth – SHIW)

- Eurostat

- The ISTAT Household Budget Survey (from the publication I Consumi delle famiglie )

123 Figure 27. Gini coefficients, Italy (1984 – 2010)

OECD Banca d'Italia Luxembourg Income Survey ISTAT Eurostat EU-SILC from Eurostat

0.40

0.38

0.36

0.34

0.32

0.30

0.28

0.26

0.24

0.22

0.20 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010 The Bank of Italy computes two types of Gini coefficients: one across households, and the second across individuals, based on a concept of equivalent income. For the purpose of comparisons, this note relies on the SHIW coefficients for equivalent incomes per person (which are lower than the Gini coefficient across households), which are available only since 1995.

The Bank of Italy is the only institution which, at the time of writing, has calculated Gini coefficients for the income year 2010, which is estimated at 0.33, unchanged since the 2008 survey.8 The Bank of Italy also reports that the “average annual household income, net of tax and social security contributions, was €32,714 in 2010”, i.e. an average monthly income of €2,726. The Bank of Italy defines equivalent income as “a measure that takes the size and composition of households into account”. The equivalent income was estimated at “€18,914 per individual in 2010, which was 0.6 per cent lower than in 2008 in real terms.

ISTAT does not include estimates of the Gini coefficients in the ISTAT Statistics database ( http://dati.istat.it/?lang=en) but includes them in the publication I Consumi delle famiglie for the years 2005, 2006, 2007, 2008 and 2009.

Over the period 1984 – 2010, income inequalities remained pretty stable in Italy, with the Gini coefficients ranging between 0.30 and 0.35, with only exception of the period 1991-1993 where inequalities recorded a significant increase. Gini coefficients from the different databases available are all included within this corridor. More detailed patterns in different sub-periods are provided below:

- From 1984 to 1993, OECD data can be compared with the Luxembourg Income Survey. Both series follow the same trend, which is not surprising as both time-series are based on the Survey on Household Income and Wealth (SHIW) conducted by the Bank of Italy. Over this decade, inequalities remained stable except between 1991 and 1993 where Gini coefficients rose up to 0.345 according to the OECD (0.339 according to the LIS) over these two years.

- From 1995 to 2004, OECD data points can be compared with those from the Bank of Italy (Survey on Household Income and Wealth), Eurostat and the Luxembourg Income Survey. Over this period of time, inequalities remained pretty stable around 0.32.

8 http://www.bancaditalia.it/statistiche/indcamp/bilfait/boll_stat/en_suppl_06_12n.pdf

124 - From 1995 to 2000, the spread between the OECD estimates and the Eurostat’s ones widened. In 2000, the spread (0.44) was the largest between the Gini coefficient computed by Eurostat (0.290) and the one calculated by the Luxembourg Income Survey (0.334). The estimates by the OECD and the Bank on Italy were between these two vlaues.

- From 2006 onwards, OECD data can be compared with Banca d’Italia, EU-SILC data and ISTAT. OECD estimates for this sub-period coincide with the EU-SILC ones. The difference between the various series narrowed over this period, which can be explained by the increasing use of the same source of inputs. (The ISTAT and OECD series shown here are both based on EU-SILC).

Overall, Figure 1. shows a growing convergence of data over time between the different datasets

The relative steadiness of Gini coefficients over time is corroborated also by the income quintile share ratio (S80/S20), see Figure 2.. The OECD income quintile share ratio increased from 4.5 in 1984 to 5.22 in 2009, but most of this increase occurred over the years 1993-1995. For 2007 to 2009, the OECD figures are perfectly matching the EU-SILC estimates. Over this long period of time, OECD time-series can be compared with the following datasets regarding the income quintile share ratio:

- Eurostat (before and after the EU-SILC)

- ISTAT (no S80/S20 on the Household Budget Survey – Data are coming from a publication labeled I Consumi delle famiglie )

Figure 28. S80/S20 ratio, Italy (1984 – 2010)

OECD Eurostat ISTAT 7

Between 1995 and 2000, the OECD estimates of the S80/S20 ratio slightly declines to 5.5 in 2004. In 2000, this ratio as calculated by the OECD was higher than the one recorded by Eurostat (5.5% for the OECD vs. 4.8% for Eurostat).

Since 2003, the OECD levels of this ratio match those computed by Eurostat based on EU-SILC. Over this period, income distribution inequality appears to have been slightly declining to 5.2% (both for the OECD and for EU-SILC) in 2009, the latest available data.

The convergence between OECD and EU-SILC time-series regarding the income quintile share ratio is also confirmed by ISTAT figures.

125 2.1.2 Time series of poverty rates

Time series on relative poverty rates based on a 50% threshold are estimated by a large number of institutions which enable comparisons over a long period of time (1984 - 2011). The below chart is showing the point-lines on poverty rates as per estimated by the OECD, the Luxembourg Income Study, the Bank of Italy, ISTAT and Eurostat (before and after the launch of EU-SILC survey)

Figure 29. Relative Poverty Rates with a 50% Threshold, Italy (1984 – 2011)

OECD Luxembourg Income Survey ISTAT Banca d'Italia Eurostat

16%

15%

14%

13%

12%

11%

10%

1984 1985 1989 1990 1991 1996 1997 1998 2002 2003 2004 2009 2010 2011 1986 1987 1988 1992 1993 1994 1995 1999 2000 2001 2005 2006 2007 2008

The latest available relative poverty rate was estimated by ISTAT at 11.1% in 2011 (absolute poverty is 5.2%)9, broadly unchanged from the previous year (11.0%). According to the national statistical office, the relative poverty threshold for a two-member household was equal to 1.011 euros per month, up from 992 euros in 2010 (without taking into consideration the variations both of consumer prices and of the mean of consumption expenditure)10.

Since 1984, relative poverty rates have been quite stable with a few movements up to now. OECD time-series can be compared over different sub-periods:

- From 1984 to 1995, relative poverty rates remained stable up to 1991 before increasing significantly between 1991 and 1994, from 11.1% to 14.2%. OECD time-series and the Luxembourg Income Survey ones perfectly match, as both time-series are based on data from SHIW. Bank of Italy’s poverty estimates are based on the concept of equivalent income.

- From 1995 to 2006, the OECD time-series shows a slight decrease of poverty rates from 14.7% in 1995 to 11.8% in 2004. This decrease is also confirmed by Eurostat, whose estimate decline from 14% in 1994 to 12.4% in 2006. Over the same period, data from Bank of Italy, ISTAT and LIS also recorded a decline, but smaller than the OECD one.

- From 2006 onwards, the OECD time-series is pretty unchanged, from 12.5% in 2006 to 12.1% in 2009, as are the Eurostat and ISTAT time-series. In contrast, the data from the Bank of Italy show a slight increase over the period (to a level of 14.4% in 2010).

9 http://www.istat.it/en/archive/67035 10 ISTAT, « Poverty in Italy 2011», 17 July 2012 in http://www.istat.it/en/archive/67035 (Full Text).

126 Estimates of relative poverty rates based on a 60% threshold follow the same trend than those based on a 50% threshold. The Figure below compare the OECD time-series to those available from the Luxembourg Income Survey and the EU-SILC data over the period 1984-2009. The 2008 value of the OECD series based on the micro-simulation model MASTRICT is shown for information.

Figure 30. Relative Poverty Rates with a 60% Threshold, Italy (1984 – 2010)

OECD Luxembourg Income Survey EU-SILC from Eurostat 25%

20%

15%

10%

Over the period 1984 - 2009, poverty rates based on a 60% threshold slightly increased, moving in a range between 16 and 22%. In 2009, the latest available data, the OECD estimated that 19% of the Italian population was living with less than 60% of the equivalised median income, a value which is fairly close than the one from Eurostat (18.2%) for the same year.

From 1984 to 1993, relative poverty rates based on this threshold increased from 16.1% to 21.9% according to the OECD. This increase is also confirmed by the Luxembourg Income Study

Between 1995 and 2004, these poverty rates slightly decreased. This decrease is more significant for the OECD data which moved down to 18.6% in 2004. The Luxembourg Income Study shows a similar decline but with a smaller variation: according to LIS, 20.3% of the Italian population was living with less than 60% of the median income in 2004, well above the OECD estimate. The OECD values based on ITAXMOD95 are higher than the ones based on MASTRICT (19.7%). In 2004, Eurostat started publishing data for Italy, showing values very close to the OECD ones (18.9% for Eurostat, compared to 18.6% for the OECD).

From 2006 onwards, the OECD dataset is based on the EU-SILC and this time-series can be compared with Eurostat estimates. The two lines are perfectly matching and the different between the two lines is always below 0.5 percentage points.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

127 Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2009 natcur 9424.25 3625.7 13.98 13,064 994.224 5766.705127 6951.94 -6646.3 20130.57144 % av HDI 46.8% 64.9% 4.9% 28.6% 34.5% -33.0%

Figure 5 compares trends in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends from 1999 onwards with the exception of the slight decrease from 2005 to 2005 which is not observable in the OECD SOCX series

Figure 5. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

No major inconsistancies can be highlighted. The following points should be considered:

Differences between the OECD reference series and the ISTAT reported data

- Gini coefficients and S80/S20 ratios as estimated by ISTAT are not available on-line; ISTAT is not publishinhg its methodology in English;

- Even if the figures are matching pretty well, the definition of relative poverty rates are different between the OECD and ISTAT. Indeed, poverty rates are based on median disposal incomes for the OECD whereas these rates are based on a average consumption expenditures for ISTAT. This term is defined as the following by the two institutions:

o The ISTAT estimate “is based on a poverty line (International standard of poverty Line- ISPL) defining as poor a household of two components with a consumption expenditure level

128 lower or equal to the mean per-capita consumption expenditure (for different size households an equivalence scale is used to take into account different needs and the economies/diseconomies of scale that can be achieved in bigger/smaller households). Therefore, the poverty line set the consumption expenditure level which represents the threshold discriminating between poor and non poor households” (ISTAT Statistics Website). The figures on relative poverty rates provided by ISTAT are labelled as “household relative poverty incidence in percentage”.

o For the OECD, “the relative poverty threshold is expressed as a given percentage of the median disposable income, expressed in nominal terms (current prices). Therefore, this threshold changes over time, as the median income changes over time. Two relative poverty thresholds are used: the first one is set at 50% of the median equivalised disposable income of the entire population, the second one is set at 60% of that income11”.

Differences between the OECD reference series and Eurostat

- The OECD time-series has been based on EU-SILC since its implementation in Italy. Therefore, Gini coefficients and S80/S20 ratios as published by Eurotstat are very close to the ones based on the OECD database since 2003 (income year);

- OECD and EU-SILC data present minor differences in relative poverty rates, which are always lower for EU-SILC than for the OECD. These differencs might be explained by the use of different equivalence scales (the OECD reference series uses the square root of household size, whereas Eurostat uses the OECD modified equivalence scale).

- Over 1994 to 2003, Eurostat computed income inequality and poverty indicators based on ECHP. Despite the different sources, levels and trends were pretty similar with OECD ones.

Differences between the OECD reference series and the Luxembourg Income Survey

- For the indicators mentioned in this Quality Review, the OECD time-series and the Luxembourg Income Survey match pretty well. This is explained by the fact that both institutions largely rely on the Bank of Italy’s SHIW. When there are discrepancies between the OECD and the LIS before 2004, they can be explained by the interpolation of data that the OECD undertook after 2006 to have a consistent time-series over time. For example, the Gini coefficient in 1991 as calculated by the OECD was estimated at 0.29 before interpolation (labeled as “previous” in the OECD database) and at 0.27 after the interpolation. The Gini coefficient recorded in the LIS for 1991 was also 0.29.

Differences between the OECD and the SHIW series from the Bank of Italy

- The OECD time-series on income inequalities and poverty rates were based on the Survey of Household Income and Wealth from 1984 to 2009 when EU-SILC was adopted.

- Gini coefficients have been always slightly higher for the bank of Italy than for the OECD. The gap between the two time-series widened over time, to 0.012 in 2008.

- Relative poverty rates have always been slightly higher for the Bank of Italy than for the OECD (between 1 and .2 percentage points). However, trends in the two series are pretty similar, with both series showing broad stability.

11 OECD Terms of References – Wave 6

129 5. Summary evaluation

Compared to some other OECD countries, long OECD time-series are available for Italy on income inequality distribution and poverty. In addition, several national statistics exist in Italy on these indicators, which enable a broad range of comparisons.

Over time and taking into account the different available databases, the OECD time-series match pretty well those used in national and European reporting. Despite slight differences regarding the levels of the different indicators, general patterns and trends are quite similar. Differences in methodology and definitions among the sources may arise and explain the observed differences in levels for all the indicators which are studied in this Review.

130 INCOME DISTRIBUTION DATA REVIEW - JAPAN

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD database relies on data from the Comprehensive Survey of Living Conditions of the People on Health and Welfare, conducted every three years by the Japanese Ministry of Health, Labour and Welfare. Estimates, computed by researchers at the National Institute of Population and Social Security Research (IPSS), are currently available for the years 1985, 1995, 2000, 2003 and 2006.

Data referring to individuals with an income three times larger than the standard deviation were excluded by IPSS before 1995. Since that year onwards, IPSS have reintegrated them in the calculations. To correct for the break in the series available in the OECD database, multiplicative adjustments (based on the 1995 ratios between the top-coded and non-top coded records) has been applied by the OECD Secretariat to the estimates provided for 1985.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Production of National Statistics in Japan is highly decentralized, with many agencies producing statistics on similar topics. The main sources available are the following:

 The Comprehensive Survey of Living Conditions (CSLC) is conducted by the Ministry of Health, Labour and Welfare every three years based on a large sample (around 30,000 households) and every year for a smaller sample. CSLC covers all private households and has a large sample size.

 The Survey on the Redistribution of Income (SRI), also conducted by the Ministry of Health, Labour and Welfare every three years, is based on a sub-sample of CSLC. Results from the 1999 wave of IRS (containing data up to 1996) are available on-line, while other publications provide more up-to-date estimates based on the same source. 12

 The National Survey of Family Income and Expenditure (NSFIE) is conducted by the Japanese Statistical Office every five years, and was used by the OECD in the past. The NSFIE sample size is the largest among all surveys (60 000 households) and the response rate is close to 100%. Some categories of households are however excluded from NSFIE (Table 1). NSFIE respondents are asked to complete several questionnaires pertaining to household income and expenditures, reasons for purchasing goods, type of outlets used for purchasing goods, major durable goods purchased, characteristic of households and their members, housing, savings and financial liabilities. Some income items (social transfers other than pensions, income taxes, residence tax, property tax, social insurance premia and occasional incomes) are however excluded. In addition, income questions are asked only to workers’ households and households whose head is unemployed. Executives of companies or corporations, which are not included among workers’ households but in other households, are hence excluded from respondents providing information on income and expenditures.

12 Fukawa T. (2006), “Income distribution in Japan based on IRS 1987-2002”, The Japanese Journal of Social Security Policy, Vol.5, No.1, June.

131  The Family Income and Expenditure Survey (FIES) is run monthly by the Japanese Statistical Office. However, its (small) sample makes this survey not representative of the entire population.

 The Keio Household Panel Survey Data (KHPS) which is conducted by Keio University since 2004, is used by the Luxembourg income Study Database (LWS). Its sample is not representative of the entire population; as data of this survey are not accessible on-line, this survey is not further considered in this review.

Table 1 presents the main characteristics of the different sources:

132 Table 21. Characteristics of datasets used for income reporting, Japan

Name Comprehensive Survey of Living Survey on the Family Income and National Survey of Family Income Keio Household Panel Survey Conditions Redistribution of Expenditure Survey and Expenditure Data (KHPS) Income Responsible Statistics and Information Department, Statistics and Information Statistics Bureau Statistics Bureau Keio University agency Minister's Secretariat, Ministry of Health, Department, Minister's Labour and Welfare. Secretariat, Ministry of Health, Labour and Welfare. Data Since 1986 Since 1972 Since 1959 (2009 latest available Since 2004 but changes of availability year, 11th edition) name and features in 2009 Frequency Large-scale survey every three years and Every three years Every month Every five years Every year small-scale survey in each interim year of large-scale survey Covered All households and household members All households and their The survey unit is the The households to be surveyed Men and women aged below 20 population nationwide. Excluded are business members nationwide. household except were the ones selected by the and above 69 years are bachelor, migrant worker, people absent The following persons institutional households method designated by the Minister excluded from the survey for extended business trips (3 months or are excluded: Live-in and one-person for Internal Affairs and (representing more than 30% of more), student living overseas, person single member households of student in Communications for all the the total population in 2004) living in a social welfare institution, long- households and those the entire area of Japan. households in the whole country. term inpatient (and whose resident living in boarding houses; The following households The survey is undertaken separately registrations are transferred to the people living in social are, however, excluded for two-or-more-person households hospital), boarded or foster child, prisoner welfare facilities as inappropriate and one-person households. The and other persons living apart from households. following households are excluded: households a. Households which (1) For two-or-more-person manage restaurants, households: a. Households running hotels, boarding houses restaurants or inns on the same or dormitories, sharing premise; b. Households running their dwellings. boarding houses, or households b. Households which with boarders; c. Households with serve meals to the four or more live-in employees; and boarders even though not d. Foreigners' households. managing boarding houses as an occupation. (2) For one-person households c. Households with 4 or a. Persons under 15 years of age; more living-in employees. b. One-person households d. Households whose corresponding to a), b) and d) of (1) heads are absent for a above; c. One-person households long time(three months or residing with live-in employees; d. more). Students; e. Inmates of social and e. Foreigner households reform institutions; f. Inpatients in hospitals and sanatoriums. Sample size Large-scale survey 2007, the In 1996 - 10000 On average, approx. On average, approx. 60 000 Approx 4 000 questionnaire covered all households households among the 9,000 households (for households. categories (approximately 290,000 households targeted in 2005, around 8,000 two- households, and 760,000 persons), the Comprehensive or-more-person For 2004, around 54,000 two-or- 133 randomly sampled in 5440 districts from Survey of the Living households and 745 one- more-person households and 5,000 the National Census in 2005. Conditions of People on person households) one-person households Small-scale survey 2008, questionnaire Health and Welfare covered all households types In 2002 10125 (approximately 58,000 households and households 150,000 persons), randomly sampled in On average, approx. 15 1,088 districts from the National Census 000 in 2005.

Sampling Stratified random sampling from the sub Same than the Three-stage stratified Households are selected separately Two-stage stratified random method district number 1 and 8 of the 2005 Comprehensive Survey sampling method. The for two-or-more person households sampling method (1st stage: Population Census’s enumeration district. of Living Conditions sampling units at three and one-person household. survey area; 2nd stage: The respondent himself/herself filled out (subset) stages are namely, There is first a selection of sample individuals) the questionnaire which was distributed primarily the municipality cities, then on unit areas and finally by an enumerator in advance, and the (i.e. city, town and a selection of sample households enumerator collected the questionnaire at village), secondly the a later date. survey unit area and thirdly the household Disseminatio Released immediately June of the following year Unknown n frequency after (annual report) compilation Sampling Household and household members One-person household and two-or- Individuals and households unit more person households levels Remark Subset of the Some households and some types Data are not in free-access. Comprehensive Survey of incomes are excluded Application to obtain access of Living Conditions should be done by post or by hand delivery. Access restricted to researchers and graduate students affiliated with national, public, and private research institutions only for nonprofit and academic purposes. Websource http://www.mhlw.go.jp/english/database/d http://www.mhlw.go.jp/en http://www.stat.go.jp/engli http://www.stat.go.jp/english/data/ze http://www.pdrc.keio.ac.jp/en/op b-hss/cslc.html glish/wp/wp- sh/data/kakei/1560.htm nsho/2009/cgaiyo.htm en/use.html hw/vol1/p1c2s2.html http://www.stat.go.jp/engli http://www.lisdatacenter.org/wp- sh/data/kakei/pdf/p2.pdf# content/uploads/2011/02/jp03sur page=8 vey.pdf

134 2. Comparison of main results derived from OECD and alternative sources

2.1 Income

2.1.1. Time series of Gini coefficients and other inequality indicators

According to the OECD estimates, the Gini coefficient for disposable income rose in Japan from 0.28 in 1985 to 0.33 in 2006. The rise is large compared to the other OECD countries. The upward trend for the Gini in the OECD database is confirmed by other national sources (Figure 2).

There are significant differences in levels of the Gini coefficients according to the different surveys. OECD estimates are lower than national estimates based on either CSLC or IRS (which are very similar) but higher that those based on NSFIE. Pseudo-Gini’s coefficients from NSFIE (which measure disparities between income groups by arranging income in order and applying the same computation method as Gini’s coefficient) are closer to the values of the Gini coefficients available in the OECD database.

Figure 31. Gini coefficients (1985 – 2009)

Differences in inequality measures between OECD and national sources reflect differences in income concepts, units of analysis, equivalisation methods, etc. These differences can also be explained by the characteristics of the different surveys:

 NSFIE provides Gini coefficients according to the characteristics of households and the age of household heads. The published Gini coefficients from NSFIE refer to workers’ households only. In 2004, the Gini coefficient for workers households was 0.257. In terms of age of the household head, Gini coefficients based on NSFIE ranged between 0.204 (35-39 years old) and 0.279 (60- 64 years old). The fact that only workers’ households are considered by NSFIE can explain the lower levels of the Gini in NSFIE that in the OECD source (e.g. CEOs, who earn some of the highest incomes, are not considered by these Gini coefficients. Table 1 provides more details on the households excluded in this survey. In addition, some incomes sources are excluded; the

135 NSFIE survey covers only the regular wages and salaries of households, and excludes other public transfers than pensions, income taxes, residence tax, property tax, social insurance premia and occasional incomes.

 National estimates of Gini coefficients based on both CSLC and SRI are very similar. Gini coefficients from both surveys are higher than those based on SFIE as all types of private households are included. CSLC has a claim to be more representative of Japanese society, although Ballas et al.13 argue that the inclusion in CSLC of students’ households and the over- representation of elderly people may lead to over-estimates of income inequality.

The 2006 Annual Report from the Japanese Economy and Public Finance also highlights a moderate increase in Gini coefficients based on various surveys (Annual Report on The Japanese Economy and Public Finance - 2006)14. According to this report, Gini coefficients from SRI/CSLS are higher than those from NSFIE due to a combination of differences in sample size and sampling methods (Table 1). In addition, the inclusion in CSLC of students living alone and of a larger number of households with income below 2 million yen (approximately 19% in 2004, as compared 10% of total households in the NSFIE) also contributes to explain the differences between the two surveys.

The Figure below shows the differences between Gini coefficients according to the various statistical surveys' data (Annual Report on The Japanese Economy and Public Finance - 2006).

Figure 32. Gini coefficients measured by various statistical surveys’ data, Japan (1985 – 2009)

In the 2006 Annual Report quoted above, the Cabinet Office Government of Japan analysed the factors contributing to the rise in the Gini coefficient for all households from 1985 to 2004 based on data

13 Ballas, Dorling, Nakaya, Tunstall and Hanaoka, “Income inequalities in Britain and Japan: a comparative study of two island economies”, draft. 14 Annual Report on the Japanese Economy and Public Finance, 2006 - Japanese Economy Heading for New Growth Era with Conditions for Growth Restored - Cabinet Office Government of Japan - http://www5.cao.go.jp/zenbun/wp-e/wp-je06/06-00303.html

136 from SRI. The study suggested that aging and other demographic effects constituted the largest factor, accounting for over 60% of the observed increase in the Gini coefficient. Jones (2007) also underscores the importance of population ageing, while also noting the growing importance of non-regular workers, who are paid significantly less than regular workers.15 Figure 4 shows that the Gini coefficient for the aged population before taxes and transfers rose significantly from 1985 to 2006, with transfers and taxes partially offsetting this rise.

Figure 33. Gini coefficients among the elderly, 1985 – 2011

Evidence on income distribution cannot be easily compared across national sources because it often refers to different concepts. National studies based on CSLC often refer to the frequency distribution by income groups. In the 2009, CSLS median income was 4,270,000 yen and the percentage of households below average income was 61.5%. These measures cannot be compared to the OECD ones, which are available for 2006 and refer to disposal incomes. National studies based on CSLC typically refer to trends in average income per household and per household member for quintile groups. These measures are significantly lower than the OECD measures but are also based on different concepts: national estimates based on CSLC refer to the S80/S20 ratio based on average income per household member, whereas the OECD ratios refer to equivalised disposal income. NSFIE estimates of income inequality are intermediate between those available in the OECD database and those reported in national studies based on CSLC. As already mentioned, NSFIE excludes some income items such as pocket money and “remittances.

15 Jones, R. S. (2007), "Income Inequality, Poverty and Social Spending in Japan", OECD Economics Department Working Papers, No. 556, OECD publishing, Paris.

137

Figure 34. S80/S20, Japan (1985 – 2010)

2.1.2 Poverty rates and composition

The increase in income inequalities in Japan was accompanied by a rise in the relative poverty rate. According to the OECD estimates, the share of the Japanese population living with less than 50% of the median equivalised income increased from 12% in 1985 to close to 16% in 2006 (last available figure). Over the same period, the poverty rate based on a 60% threshold increased from 17.8% to 21.7%, an increase that is significantly % higher than in other OECD countries.16 In 2004, the relative poverty rates of Japan was the fourth highest among all OECD countries, while that for households with a head of working age with children and one adult households was the highest.17 According to some analysis, this rise in the relative poverty rate can be explained by the population ageing and the increase in the share of single- person households. 18 The results of the OECD are very similar to those highlighted by national reporting based on CSLC

16 Mira d’Ercole (2006), “Income Inequality and Poverty in OECD countries: how does Japan to compare?”, The Japanese Journal of Social Security Policy, Vol. 5, No.1, June. 17 OECD (2008), Growing Unequal? Income distribution and Poverty” http://www.mhlw.go.jp/english/wp/wp- hw4/dl/honbun/2_2_3.pdf 18 Jones, R. S. (2007), "Income Inequality, Poverty and Social Spending in Japan", OECD Economics Department , Working Papers, No. 556, OECD publishing, © OECD.

138 Figure 35. Trends in poverty rates, based on a threshold set at 50% of median income

Figure 36. Trends in poverty rates, based on a threshold set at 60% of median income

As in the case of income inequality, poverty rates based on NSFIE are lower than both the OECD and national estimates based on CSLC. As noted above, this can be explained by the fact that not all the private households are included in NSFIE and that some sources of incomes are excluded.19 National reports based FIES and the SRI do not provide figures on poverty rates.

Regarding child poverty rates, OECD and national estimates from CSLC are very similar, although data cannot be compared before 1997. In both cases, child poverty rates increased since 1985.

19 Ballas et al. estimated poverty rates based on gross income figures to cope with the lack of available data on disposable income; these are lower than the ones mentioned above.

139 Figure 37. Trends in Child Poverty Rates (at 50% median income)

Poverty rates for single adults (of working age) with children reported in the OECD database and in national reports based on CSLC are very similar, even if data cannot be compared before 1997. These rates recorded a significant increase from 1985 to 1997 before diminishing afterwards. The poverty rate for this type of households is particularly high in Japan.

Figure 38. Trends in relative poverty rates for people living in single adult households with children

140 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Such information is not available for the other data sources described in table 1.

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery OECD reference suvery 2006 natcur 1714547 336245 452293 2,503,086 170068 323833.2328 628106 -662082.7 2963009.492 % av HDI 57.9% 84.5% 5.7% 10.9% 21.2% -22.3%

Figure 9 compares shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show a rising trend throughout the period, sharper in the OECD income distribution database than in OECD SOCX. Differences in levels of the two series are relatively small in the mid-1990s, but significantly in the mid-2000s.

Figure 9 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Overall, significant differences between the different data sources are evident in terms of definitions, methodology and data treatment.

 For Gini coefficients, OECD estimates are intermediate between the four national sources. As described above, the OECD relies on data from CSLC, based on the view that this gives a more accurate and complete picture of income inequalities.

 For all measures of income distribution, the OECD refers to the concept of disposal income per consumption unit whereas national estimates based on CSLC refer to average income per

141 household and per household members. Inequality estimates from NSFIE differ from those from other sources also because some income types are excluded.

 For poverty rates, the OECD results are similar to national estimates from CSLC for total households, children and lone parents. Poverty rates based on NSFIE are lower than those from other sources due to the exclusion of some types of households and incomes items.

5. Summary evaluation

Overall, OECD references series show higher income inequalities than those based on either FIES or NSFIE. While the OECD relies on CSLC as its reference source, this reflects the choice made by the Japanese authorities providing the data; the CSLC is also the data source for Japan included in the “Survey of Country Practices” undertaken as part of the 2011 Canberra Handbook on the Measurement of Household Income. While some studies (e.g. Pickett and Wilkinson “The Spirit Level”) have argued that Japan is “a more equitable and society in terms of income than is any other industrialised country”, this conclusion depends on the use of an alternative official data source (NSFIE) that appears as less suited for assessing income inequality and poverty. Better understanding the full range of methodological differences between the different national sources would require further investigation, as most methodological documents available on-line are written in Japanese with only partial translation in English or French (http://www.e-stat.go.jp/SG1/estat/NewListE.do?tid=000001037021). Finally, even if the differences between the several sources had been tentatively identified in terms of households’ coverage and income items, it is difficult to estimate the number of households and the total sum of income that are excluded from one survey to another.

Bibliography

Annual Report on the Japanese Economy and Public Finance, 2006 - Japanese Economy Heading for New Growth Era with Conditions for Growth Restored - Cabinet Office Government of Japan - http://www5.cao.go.jp/zenbun/wp-e/wp-je06/06-00303.html

Ballas, Dorling, Nakaya, Tunstall andHanaoka, “Income inequalities in Britain and Japan: a comparative study of two island economies”, draft.

Fukawa T. (2006), “Income distribution in Japan based on IRS 1987-2002”, The Japanese Journal of Social Security Policy, Vol.5, No.1, June.

Jones, R. S. (2007), "Income Inequality, Poverty and Social Spending in Japan", OECD Economics Department , Working Papers, No. 556, OECD publishing, © OECD.

Mira d’Ercole, M. (2006), “Income Inequality and Poverty in OECD countries: how does Japan to compare?”, The Japanese Journal of Social Security Policy, Vol. 5, No.1, June.

OECD (2008), Growing Unequal? Income distribution and Poverty”: http://www.mhlw.go.jp/english/wp/wp-hw4/dl/honbun/2_2_3.pdf

142 INCOME DISTRIBUTION DATA REVIEW –KOREA

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Korea are computed by Statistics Korea from the Household Income & Expenditure survey combined with Farm household economy survey. Data are available annually from 2006 to 2011.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Income distribution and poverty indicators for Korea are also available from:

 The same combination of Household Income & Expenditure survey combined with Farm household economy survey from previous years, using different coverage of population

o Since 2006: all households (same series as OECD reference series)

o Since 2003: only for 2 and more non-farm household

o Since 1990: only for urban salary and wage worker, 2 and more household

 KIHASA’s series from KOWEPS(Korea Welfare Panel Study), every year from 2005 to 2009

 Two series based from the Korea Labor Institute ‘s KLIPS (Korean Labor & Income Panel Study) published in (1) Jang & Lee (2010) and (2) Kim (2011), every year from 1997 to 2006/7.

1.2.2 International reporting:

OECD earnings indicators for Korea are from Entreprise Survey (Wage Structure Survey) from the Korean Ministry of Labour (Yearbook of Labour Statistics). (see Part II).

Table 1 presents the main characteristics of the different sources:

143 Table 1. Characteristics of datasets used for income reporting, Korea

National survey (Income)

Name Combination of Household KOWEPS (Korea Welfare KLIPS (Korean Labor & Income Panel Income & Expenditure survey Panel Study) Study) combined with Farm household economy survey Name of the responsible Statistics Korea, Welfare Korea Institute for Health and Korean Labor Institute(Now changed into agency Statistics division Social Affairs(KIHASA) Korean Employment Information Service)

Year (survey and HIES (2011:Q2) and FHES 2005 to 2009 income, every 1997 to 2006/7, every year income/wage) (2011:Q4) representing 2010 year income Period over which income is Annual income for the all year Annual income for the all year Annual income for the all year N-3 assessed N-1 N-2

Covered population from 2006 onwards: all All households including urban All households living in urban areas population/households; and suburb area across the country from 2003 onwards: only for 2 and more nonfarm households; from 1990 onwards: only for urban salary and wage worker, 2 and more households Sample size 999 entity districts per month 6034 households(5th wave) 5379 households representing 8800 households

Sample procedure Cross-sectional survey. stratified cluster sampling two-stage stratified clustering sampling. Rolling system(1/3 of sample (Over-sampling of people households are changed under median income 60% every year); once a household line) is selected in a sample, Income and expenditure data are collected for about 3 years.

Response rate 81% N/A N/A Imputation of missing values Yes yes Yes Unit for data collection Household and individual households and individuals households and individuals living in urban living in both urban and suburb areas. areas. Break in series - - -

Web source: http://www.moel.go.kr/english/stat (1) (1): Jang & Lee(2010).An Analysis on the istics/MOL_Survey_Data.jsp http://koweps.re.kr/0903bbs/bbsd Korean Household Income Inequality by own.asp?type2=3&num=145 Using the Gini Decomposition. (2)http://eng.keis.or.kr/eng/board/ APEA2010. Sixth Annual Conference. NR_boardView.do?bbsCd=1023& (2): Kim(2011). The Dynamics of Income seq=20120704161021764¤t Page=&ctgCd=&sortName=&sort Inequality in Korea,The center of regional Order=&startDt=&endDt=&pageTy and global economic studies at Bryant pe=®Pwd=N&showSummaryY University working paper series. n=&startDtVal=&endDtVal=&sear chKey=TITLE___1002&searchVal =

144 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to both the OECD income distribution database and the Korean source from 2006, income inequality among total population was rather stable between 2006 and 2011 at around 0.31, which is also the level of the OECD average in 2008. Using the same survey and the same methodology, it is not possible to observe long-term trends of income inequality. Data are however available from 2003 for 2 and more non-farm households and from 1990 for households of two persons or more for urban salary and wage workers. The Gini coefficient is 0.02 points lower in these longer series, representing slightly lower inequality within a more restricted sample of the population. Inequality would have gradually increased from 0.25 in 1990 to to 0.29 in 2011.

KIHASA’s series from the Korean Welfare Panel Study (KOWEPS) shows a declining trend from 0.335 in 2005 to 0.31 in 2009.

KLIPS series from (1) Jang & Lee (2010) and (2) Kim (2011) show a similar gradual increase in inequality from 1997 to 2006/7, but a a higher level of inequality of 0.40.

Figure 1.1 Trends in Gini coefficient (disposable income)

The inter-decile ratio (P90/P10) from the OECD reference survey is slightly higher than the one from the Poverty Statistics Yearbook from Kihasa (http://www.kihasa.re.kr/html/jsp/publication/research/view.jsp?bid=12&ano=1194). Both series show a gradual increase in inequality from 2006 to 2009, then the ratio is stable in 2010 and 2011.

145 Figure 1.2 Trends in P90/P10

2.1.2 Time series of poverty rates

According to both the OECD income distribution database and the Korean source from 2006, the share of the Korean population living with less than 50% of the median equivalised income (9 994 000 Won per year in 2011) has increased slightly from 14.3% in 2006 to 15.3% in 2009, then it slightly declined to 14.9% in 2010 and back to 15.2% in 2011.

The OECD reference series is 2-3 % point higher than longer series from the combined HIES & FHES with a more restricted sample of the population, showing a gradual increase in relative poverty rates from 7% in 1990 to around 12% in the late 2000s. We can also see a relatively strong increase in poverty in 1998 and 1999 during the Asian crisis. The level went back to 9% in 2000.

KIHASA’s series from the Korean Welfare Panel Study (KOWEPS) shows higher levels of poverty and a strong declining trend from 16.4% in 2005 to 13.5% in 2009.

Figure 2.1 Trends in poverty rates

146 As for child poverty, the OECD reference series shows a continuous downward trend from 10.9% in 2007 to 9.4% in 2010, then up to 9.7% in 2011. The Korean source from 2006 shows slightly lower rates but a similar trend.

Figure 2.2 Trends in Child poverty rates

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for 2011 (the latest available year), according to the OECD benchmark series. Unfortunately, such information could not easily be found for the other data sources described in table 1.

Table 2. Shares of income components in total disposable income, OECD reference series

Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show upward trends throughout the period, except for the latest year where SOCX data suggest a slight decline.

147 Figure 3 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

The OECD reference series fully matches the Korean series from the same data source from 2006. There are no differences. The main difference with series from the same survey but from a different starting year come from the population covered in the survey. The OECD reference series covers the total population, whereas the series from 2003 only covers only for 2 and more nonfarm households, and the series from 1990 only covers urban salary and wage worker, 2 and more households. This would explain the lower inequality and lower poverty rates in the two alternative series.

KIHASA’s series from KOWEPS and KLI’s series from KLIPS show different trends which are still to be explained.

5. Summary evaluation

The OECD reference series fully matches the series published by Statistics Korea on the basis of the same data source (combined HIES & FHES from 2006). There are no differences The OECD reference series shows (still unexplained) lower levels of inequality than the KLIPS series but it shows similar gradual increase trends in inequality. On the other hand, KIHASA’s series from the Korean Welfare Panel Study (KOWEPS) shows more different trends, sometimes in the opposite direction than those suggested by the former series.

As there are data for 2006 in both series from 2006 and from 1990, the 2006 series could be trended backward to compute estimates for longer-term trends for Korea from 1990.

148 INCOME DISTRIBUTION DATA REVIEW – LUXEMBOURG

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Luxembourg are computed by Statec (Statistical Office of Luxembourg) and are based on two surveys, for the following years available in the OECD database:

 The Panel Socio Economic "Liewen Zu Letzebuerg”, for the years 1986, 1996, 2001 and 2004

 EU-SILC, for the years 2008, 2009

EU-SILC having been developed on the basis of the Panel Socio Economic, there is no break between the series.

1.2. National reporting and reporting in other international agencies:

 Eurostat’s EU-SILC annual survey since 2000 (2001 missing).

 LIS database, using annual surveys located at CEPS/Instead.

 The Statistical Office of the Grand-Duchee of Luxembourg (edited by STATEC) using annual surveys from 2003 to 2011.

The below table presents the main characteristics of those four datasets:

149 Table 22. Characteristics of datasets used for income reporting, Luxembourg

OECD reference series National survey Eurostat LIS database income distribution (Income) database Name EU-SILC STATEC EU-SILC ECHP (1994, 1997, 2000) Panel Socio Economic "Liewen Zu Letzebuerg" PSELL 3 / EU SILC 2005 Name of the Eurostat EU-SILC Eurostat CEPS/Instead responsible agency Year (survey 1995, 2000-2009. 2003-2011 survey 2000-2001 and 2003- 1991 (PSELL), 1994, 1997, and representing 2002-2010 2010 survey representing 2000 (ECHP), 2004 (SILC) income/wage) income years 1999-2000 and 2002- income years. 2009 income years. Period over Annual income in the Annual income for the Annual income for the all Annual income for the all which income is previous year, also in all year N-1 year N-1 year assessed case of transfers from public sources Covered The stratification is The stratification is The stratification is done Persons living in private population done by household size, done by household size, by household size, status households status of activity (active, status of activity (active, of activity (active, retired) retired) and sector of retired) and sector of and sector of activity activity (public, private, activity (public, private, (public, private, self- self-employed) into 18 self-employed) into 18 employed) into 18 strata. strata. strata. Sample size Actual sample size: Actual sample size: Actual sample size: 8983. 3622 households containing 8983. Achieved sample 8983. Achieved sample Achieved sample size: 9661 individuals (2005). size: 4876 (2010 size: 4876 (2010 4876 (2010 survey) survey) survey) Sample Stratified simple Stratified simple random Stratified simple random Stratified sampling procedure random sampling sampling sampling Response rate 56% 56% 56% 57% Imputation of No missing values, Imputation of item non- missing values negative values treated response has been carried as suggested in the out. Four generic models terms of references. based on the « Imputation and Variance Estimation » developed by the University of Michigan. Next to these models, specific procedures (simulation, regressions, deductions) have been used for certain income components. Unit for data Fiscal Households Fiscal Households Household and invidual collection level information

Break in series No No No No

Web source: http://stats.oecd.org/Ind H:\INCOME_DISTRIBU http://epp.eurostat.ec.euro http://www.lisdatacenter.org

ex.aspx?QueryId=2606 TION\Quality pa.eu/portal/page/portal/in /techdoc/fr/frindex.htm 8 assessment\ELS\LUX- come_social_inclusion_livi ELS\Affichage de ng_conditions/quality/nati

tableau - Indicateurs onal_quality_reports d'inégalité dans la répartition des revenus 2003 - 2011.mht

150 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality in Luxembourg is lower than the OECD average (0.288 in the late 2000s against 0.314 for the OECD average). However, while income inequality in Luxembourg remained largely steady at 0.259 up until 1995, it has since then seen an increase reaching 0.29 in 2008, followed by a decline in 2009.

The other three series of Gini coefficients on disposable income in Luxembourg reveal similar levels and trends as the OECD series. The EU-SILC series and the STATEC series are identical (STATEC uses same methodology and data as EU-SILC).

Figure 39.1 Trends in Gini coefficient (disposable income)

OECD reference series Eurostat (EU-SILC) LIS (CEPS/Instead) STATEC (EU-SILC) 0.30

0.28

0.26

0.24

Gini (at disposable income) (at disposable Gini 0.22

0.20 1990 1995 2000 2005 2010

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference series, the EU-SILC series, and the STATEC series, trends between EU-SILC and STATEC are identical and show fluctuations since 2003. The OECD reference series, while constant until 2004, shows an increase thereafter reaching similar levels in 2009 as the EU-SILC and STATEC series (4.0 for the OECD series and 4.1 for the EU-SILC/STATEC series).

151 Figure 1.2 S80/S20

OECD reference series Eurostat (EU-SILC) STATEC (EU-SILC) 5.0

4.5

4.0 S80/S20

3.5

3.0 1990 1995 2000 2005 2010

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Luxembourg population living with less than 50% of the median equivalised income (20027 Euros per year) has increased from 5.5% in 1995 to 8.1% in 2009.

The EU-SILC series and LIS series show similar and consistent trends as they both show increasing levels of poverty rates since 2000, although the LIS series shows slightly higher levels than the OECD series, while the EU-SILC series shows slightly lower levels than the OECD series.

Figure 2.1 Trends in poverty rates

152 As for child poverty, the OECD reference series and EU-SILC series also show similar and consistent trends, although the EU-SILC shows slightly lower levels of child poverty rates. While the LIS series shows higher levels of child poverty rates, the increase it exhibits is similar to the other series, even if data is only available until 2004.

Figure 2.2 Trends in Child poverty rates

OECD reference series Eurostat (EU-SILC) LIS (CEPS/Instead)

15%

12%

3% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 0% 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 23. Shares of income components in total disposable income, OECD reference series

Survey Year Unit Wages Capital Self Employment Transfers Taxes Disposable income OECD reference suvery 2008 natcur 34,574 1,569 2,950 12,490 -11,635 40,055 % av HDI 86% 4% 7% 31% -29%

153 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Equivalence scale: The OECD reference series (as well as the LIS series) use the square root of household size, whereas the EU-SILC series and STATEC series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

The OECD benchmark series is consistent with different series available for Luxembourg overall. The light discrepancies may be explained by the different equivalence scales used (square foot of household size for the OECD series; OECD modified equivalence scale for EU-SILC series).

154 INCOME DISTRIBUTION DATA REVIEW - MEXICO

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD has been using the same survey for Mexico since 1984: data provided by the Survey of Household Income and Expenditure Survey (Encuesta Nacional de Ingreso y Gastos de los Hogares) computed by the National Office of Statistics (Instituto Nacional de Estadistica y Geografia - INEGI). Currently, the OECD income distribution database contains data for the following (income) years: 1984, 1994, 2000, 2004 and 2008 and 2010.

Data had been revised from 2008 onwards due to a change in methodology. From 2008, household weights have been calibrated with the population and housing census 2010. However, this change had an insignificant impact on data.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

The following datasets are available for Mexico on income distribution and poverty at a national level:

 CONEVAL (Consejo Nacional de Evaluación de la Política de Desarrollo Social) is also producing data and studies on poverty for Mexico. This National Council for the Evaluation of Social Development Policy broadly aims to regulate and coordinate the evaluation of the Social Development Policy and Programs implemented by public dependencies. The information used by the CONEVAL are also generated by the National Statistics and Geography Institute (INEGI), with a two-year minimum periodicity for state information and five-year minimum periodicity for municipal disaggregation. Please, note that there are using a multi-dimensional approach on poverty that is not comparable with OECD series.

1.2.2 International reporting:

The following datasets are available for Mexico on income distribution and poverty at an international level:

 the Luxembourg Income Survey is compiling data for Mexico from the same source as the OECD, the Survey of Household Income and Expenditure Survey (Encuesta Nacional de Ingreso y Gastos de los Hogares). The LIS database contains data for the following income years: 1984, 1989, 1992, 1994, 1996, 1998, 2000, 2002 and 2004. From 1992, data have been available every two years.

Table 1 presents the main characteristics of the different sources:

155 Table 24. Characteristics of datasets used for income reporting, Mexico

Name Household Income and Expenditure Survey (Encuesta Luxembourg Income Survey Nacional de Ingresos y Gastos de los Hogares - ENIGH) Name of the National Statistical Institute (Instituto Nacional de National Statistical Institute (Instituto Nacional de responsible Estadistica, Geografia e Informatica – INEGI ) Estadistica, Geografia e Informatica – INEGI ) Goal To obtain information on the distribution, amount and structure of incomes and expenditures for the household with the final aim of evaluating the developments in the standards of living of the population Year Every other even year 1984, 1989, 1992, 1994, 1996, 1998, 2000, 2002 and 2004 Data Annually Same than ENIGH collecting frequency Covered All national or foreign households living in private Same than ENIGH population dwellings in the national territory. Households living in collective dwellings are not included in the sample. Sample size In 2004: 22,595 households including 91,738 individuals, Same than ENIGH of which 91,450 household members (the rest being domestic servants or guests). In 2001: 17,167 households including 72,602 individuals (144 domestic servants or guests). Sampling Stratified multi-phase sample with the dwelling as Same than ENIGH method primary sampling unit: first, basic geostatistical areas (AGEB), stratified according to 5 geographic and socio- economic criteria, are selected, then, for urban areas only, blocks of dwellings, and finally dwellings from each area (rural areas) or block (urban areas). Sampling unit Households and individuals Same than ENIGH Response In 2004 - 90% of the sampled households were Same than ENIGH rates successfully interviewed In 2002 - 86% of the sampled households were successfully interviewed In 2000 - 86% of the sampled households were successfully interviewed Websource http://www.inegi.org.mx/default.aspx http://www.lisdatacenter.org/our-data/lis- database/by-country/mexico-2/

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Gini coefficients for Mexico are available in the Luxembourg Income Survey (LIS), in the ENIGH (Encuesta Nacional de Ingresos y Gastos de los Hogares) and in CONEVAL databases.

The ENIGH is calculating three types of Gini coefficients which are calculated both before and after taxes and transfers:

 A Gini coefficient according to the average of the total quarterly income per household and in constant prices (Ingreso corriente total promedio trimestral por hogar – Precios constants);

156  A Gini coefficient according to the average of the total quarterly income per capita and in constant prices (Ingreso corriente total promedio trimestral per capita – Precios constants);

 A Gini coefficient according to the average of the total quarterly income per capita and according to equivalence of scales - in constant prices (Ingreso corriente total promedio trimestral per capita ajustado por economias de escala – Precios constants).

For the purposes of this study, OECD data sets are compared with the ENIGH’s Gini coefficients according to the average of total quarterly income per capita and in constant prices20.

In terms of figures, Mexico is still recording very high levels of inequalities. Inequalities increased from 1984 to 1994 up to a Gini of around 0.5 (0.519 as per the OECD and ENIGH and 0.4850 as per LIS). After 1994 onwards, Gini coefficients declined. The Luxembourg Income Survey recorded a decrease from 0.485 in 1994 down to 0.457 in 2004, and the OECD series from 0.519 to 0.473. These are significant changes. In 2010, the latest available Gini coefficient was estimated at 0.46 according to the OECD.

Note that CONEVAL published a few GINI coefficients in some of their publications based in ENIGH data. They recorded higher Gini coefficients but with a similar trend than ENIGH.

Very logically, data are very similar, even identical, between the OECD, the ENIGH and the ENIGH with OECD methodology. The LIS time-series on Gini coefficients is following a very similar trend while recording Gini coefficients slightly below the ones from the OECD ones. On average, over the period of comparison from 1984 to 2004, the spread between the OECD and LIS remained constant about 0.02.

Figure 40. Trends in Gini coefficients (at disposal income)

OECD Reference series ENIGH - Average of the total quarterly income per capita LIS ENIGH with OECD methodology CONNEVAL 0.6

0.55

0.5

0.45

0.4

0.35

The series on income inequality ratio are following the same trend than the Gini coefficients. The OECD references series showed that the S80/S20 ratio increased from 1984 to 1994 to reach a ratio of 15.5. From 1994 onwards, this ratio recorded a slight and continuous decline down to 12.7 in 2010.

20 I.e., Se refiere al Ingreso Corriente Total promedio trimestral por persona a precios constantes del 2010, http://www.inegi.org.mx/sistemas/sisept/default.aspx?t=ming10&c=22214&s=est

157 The series published by ENIGH show a very similar pattern than the OECD ones. On average, from 2000 to 2010, there is a difference of 1 percentage point between the two time-series.

Figure 41. Trends in S80/S20

OECD S80/S20 ENIGH S80/S20 18 16 14 12 10 8 6 4 2 0

Similarly than the study of the S80/S20 ratios, the OECD time-series is recording an increase of the income inequality ratio from 1984 to 1994 to 33.52%. After that date, this ratio progressively declined down to a level of 28.525% in 2010 which is still higher than the level recorded in 1984.

The series published by ENIGH show a similar pattern than the OECD ones. Data can be compared only from 2000 onwards (data were not available before that date for ENIGH). This income inequality ratio decreased more significantly for ENIGH than for the OECD as these figures moved from 34.2% in 2000 down to 23.1% in 2010 for ENIGH. For the OECD, these ratios decreased from 30.6% to 28.5% over the same time period.

Some income inequalities ratios had been found for CONEVAL but there have been found only for two years so it is not representative of a long-term trend. Still, a delinking trend in the early 2000s is confirmed.

2.1.2 Time series of poverty rates

In Mexico, CONEVAL is the institution in charge of measuring poverty. Traditionally, poverty measurement in Mexico has been developed from a one-dimensional perspective, where income is used as an approximation to the population’s economic wellbeing. Since 2008, CONEVAL adopted a new methodology for poverty measurement which is multidimensional. Poverty is measured in two basic spaces: income, ie. well-being and deprivation, ie. social rights.

According to this new conception, a person is considered in situation of multidimensional poverty when his/her income is insufficient to acquire the goods and services he/she requires to satisfy his/her needs and presents deprivation in at least one of the following six indicators: educational gap, access to healthcare, access to social security, housing quality and spaces, basic services in homes and access to food.

158 With this methodology, CONEVAL can identify the percentage of population in situation of multidimensional poverty (extreme poverty and moderate poverty) and the percentage of situation in social deprivation.

However, this unique Mexican framework prevents to undertake any comparison regarding poverty indicators with other datasets. Indeed, even if we consider only income poverty situation, the poverty line in Mexico is expressed based on the monetary value of a basic goods and services basket. According to the Poverty Measurement Methodology, the income employed for this measurement is the Total Current Net Income Per Capita (INTPC) and the poverty thresholds are defined in three levels:

 Food poverty: Incapability to acquire a basic food basket, even if the entire income available to the household were used just to buy said basket goods.

 Capabilities poverty: Insufficiency of the available income to acquire the food basket value and make the necessary expenses in health and education, even if the total household income were devoted solely to these purposes.

 Patrimony poverty: Insufficiency of the available income to acquire the food basket, as well as to make the necessary expenses in health, education, clothing, housing and transportation, even if the entire household income were used exclusively for the acquisition of these goods and services.

These three poverty thresholds are different from the relative poverty threshold defined by the OECD which is expressed as a given percentage of the median disposable income, expressed in nominal terms (current prices).

CONEVAL also calculates the percentage of population living below the minimum wellbeing which can be defined as “the monetary value of a basic food basket in a particular month. For purposes of poverty measure, the reference value of the basket is the month of August of the year when the measurement is held. This line is calculated for rural and urban areas21”. However, this figure is also different from the OECD poverty line. Data on poverty as defined by CONEVAL are available from 2008 onwards when the new methodology has been adopted.

Therefore, the OECD time-series on income poverty rates can be compared only with the Luxembourg Income Survey which is based on ENIGH figures. Thus, OECD and LIS are using the same datasets and the point’s lines are consequently very similar on poverty rates. On the three years of comparison, 1994, 2000 and 2004, data vary of 0.2 point with the exception of 1994 where the spread between the two time series was of one percentage point.

In terms of analysis, poverty rates remained stable around 20% over the all period of time from 1984 – 2010. Since 2004, poverty is rising again after registered a minor decline over the previous decade. In 2010 which is the latest available data, 20.41% of the Mexican people live with less than 50% of the median income according to the OECD.

While presenting different types of poverty calculation, CONEVAL has been recording an increase of inequalities from 2008 to 2010. As per this institute, 46% of the Mexican was living in a situation of poverty in 2010 which is representing 52 millions of individuals.

21 http://www.coneval.gob.mx/cmsconeval/rw/pages/medicion/glosario.en.do

159 Figure 42. Trends in Poverty rates, with a 50% threshold

OECD - After taxes and transfers - Pov Line of 50% CONEVAL- Population in situation of poverty LIS - After taxes and transfers - Pov Line of 50% CONEVAL - Population with income below the minimum wellbeing line 50% 45% 40% 35% 30% 25% 20% 15% 10%

Regarding the poverty line with a 60% threshold, OECD time-series on poverty rates can again be compared only with the Luxembourg Income Survey. Similarly to the poverty line with a 50% threshold, data are very similar between the OECD and LIS. Over the three years of comparison, 1994, 2000 and 2004, data vary of 0.2 point with the exception of 1994 where the spread between the two time series is of one percentage point.

From 1984 to 2000, the share of Mexican living with less than 60% of the median income rose from 25.1% to 28%. Since 2000, this figure declined down to 25% in 2004 before recording new increases until 2010. In 2010, which is the latest available data from the OECD, the poverty rate with a 60% threshold was estimated at 26.7%.

Figure 43. Trends in Poverty rates with a 60% threshold

OECD figures - After taxes and transfers - Pov Line of 60% LIS - After taxes and transfers - Pov Line of 60% 40%

30%

20%

10%

160 2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery OECD reference suvery 2010 natcur 22287.9 5572.7 13189.4 41,050 3522.3 6508.800516 5124.34 0 56205.4488 % av HDI 39.7% 73.0% 6.3% 11.6% 9.1% 0.0%

Figure 6 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period. To note that the increase of public social transfers is more significant according to the OECD reference series than to the OECD Social Expenditure database (OECD SOCX).

Figure 6. Trends in shares of public social transfers

10 % transfers in disposable income (OECD reference series) 9 % cash social spending in Net National Income (OECD SOCX) 8 7 6 5 4 3 2 1 0 1990 1995 2000 2005 2010

4. Metadata of data sources which should explain differences and inconsistencies

Regarding the income inequality ratios, the OECD time-series cannot be compared with complete validity with the LIS data due to a slightly different methodology.

Regarding the comparison on poverty rates, the data provided by CONEVAL cannot be compared with the OECD ones as this office is measuring poverty over a multidimensional approach.

161 5. Summary evaluation

Comparisons of alternative data with the OECD time-series are largely based on the same survey, the ENIGH.

Regarding the estimates for Gini coefficients, the following remarks can be pointed out:

 There is a similar trend between the OECD, the ENIGH, the LIS and CONEVAL.

 The LIS data has been recording slightly lower Gini coefficients than the OECD while presenting a consistent spread over the period time.

 Several Gini coefficients are calculated by ENIGH which often differ in levels but not in trend estimates.

Regarding the poverty rates, the LIS and the OECD are presenting very similar trend and rates as both of them are using the same survey, the ENIGH one and the same relative poverty thresholds (50% and 60% of median income, respectively).

To conclude, the OECD reference series and the LIS ones based on the ENIGH survey match well. In addition, the OECD reference series and the ENIGH reference series are presenting very similar trends and the spread between the different figures are usually very narrow.

162 INCOME DISTRIBUTION DATA REVIEW –NETHERLANDS

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD Income Distribution Data for the Netherlands are provided by Statistics Netherlands and are based on the Sociaal-echonomisch panelonderzoek/ Dutch Socio-Economic Panel (SEP): In 1984, Statistics Netherlands has started the Socio Economic Panel survey (SEP). This survey follows approximately 5 000 households through time. For this purpose, all household members aged 16 years and over are periodically interviewed about their socio economic situation with questions on education, labour market participation, income, assets and debts. In addition, one of the household members, preferably the head or the spouse/partner, is asked questions concerning living conditions, ownership of consumer goods and income evaluation for the entire household. Finally, of all household members, including children below 16 years of age, data on gender, date of birth, marital status, nationality and household situation are recorded.

In 2006 there is a break in series due to a policy change in the system of Health cost insurance. Excluding employers’ contributions to social security make figures incomparable over time.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

 Inkomenspanelonderzoek/ Income Panel Survey (IPO): The Income Panel Survey has been designed to have a picture of the composition and distribution of the income of individuals and households in the Netherlands. In contrast with general panel surveys based on personal interviews, this is a register panel with information from different administrative sources. The advantages are that there is almost no panel attrition and almost no non response (about 2%); another relies in the large number of the sample. Detailed income and wealth information is recorded by socioeconomic category and region. The IPO outcomes are published on StatLine (the Statistics Netherlands output Database) and used in the Distribution of Wealth Yearbook, the Poverty Monitor and various other articles and press releases. Statistics Netherlands uses the Income Panel Survey as reference for income inequality and poverty indicators. Potential issue: As student households and households without income throughout the year are excluded from the target population in the Income Panel Survey (IPO), the data report lower poverty and inequality indicators.

 Additional Enquiry on the Use of (Public) Services: The Additional Enquiry on the Use of (Public) Services is conducted every 4 years. The first year for which the survey was conducted was 1979, and was first made available as microdata in 1979.The main purpose of the AVO is to measure income, household composition, and the use of the following public services: education, health care, housing and social and cultural activities. Services such as public transport are not included. Potential issues: Users of this data should be aware of the following problems regarding the quality of the income data collected in this survey. 1. There are too few self-employed receiving low incomes. 2. Recipients of unemployment and disabled benefits, social assistance, and old age pensions are underrepresented. 3. Income was recorded for weekly, monthly, and annual periods; in some cases these may be incorrect.

163  Sociaal-echonomisch panelonderzoek/ Socio-Economic Panel Survey: see above as the OECD reporting series.

1.2.2 International reporting:

 Eurostat is also computing some indicators on income distribution and poverty for the Netherlands based on the EU Survey on Income and Living Conditions (EU SILC).

 The Netherlands is also included in the Luxembourg Income Study Database (LIS) using Additional Enquiry on the Use of (Public) Services till 1990, then the Socio-Economic Panel Survey (SEP) and the Survey on Income and Living Conditions/EU-SILC since 2004.

The below table presents the main characteristics of the different sources:

164 Table 1. Characteristics of datasets used for income reporting, the Netherlands

165 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Figure 1.1 Trends in Gini coefficient (disposable income)

Source: Dutch Socio-Economic Panel / Sociaal-echonomisch panelonderzoek (SEP); Statistics Netherlands: Inkomenspanelonderzoek/ Income Panel Survey (IPO); Eurostat, European Community Household Panel (ECHP), EU Survey on Income and Living Conditions, (EU SILC), LIS: Cross national data center in Luxembourg http://www.lisdatacenter.org/.

Trends in Gini coefficient are quite similar between the OECD reference series (SEP) and the data reported by Statistics Netherlands from the Income Panel Survey (IPO). However trends differ more widely for LIS data and Eurostat (EUSILC). IPO reports lower level of income inequality probably because households without income throughout the year are excluded from the target population.

Figure 1.2 Trends in S80/S20

166 2.1.2 Time series of poverty rates

Figure 2.1 Trends in poverty rates

Trends in poverty rates seem to be more in line between the OECD reference series (SEP) and EUSILC except for the last year. As student households and households without income throughout the year are excluded of the target population in the Income Panel Survey (IPO), the data report lower poverty rates.

Figure 2.2 Trends in child poverty rates

2.2 Wages

See Part II of the present Data Review.

167 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series and EU SILC. The two series match relatively well, with the exception of the share of capital incomes which have a higher share in the OECD series and transfers, which have a lower share.

Table 2. Shares of income components in total disposable income, OECD reference series and EU SILC

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference series (Income Panel Survey) 2008 natcur 17 524 5 636 1 742 24 903 4 407 3 279 4 509 -11 580 25 518 % av HDI 69% 98% 17% 13% 18% -45%

EU SILC 2008 natcur 16 910 8 392 11 25 313 1 834 2 995 7 280 -12 902 24 795 % av HDI 68% 102% 7% 12% 29% -52%

natcur 1.04 0.98 2.40 1.09 0.62 0.90 1.03 % av HDI 1.01 0.96 2.34 1.06 0.60 0.87

Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period, except for 2006 and for the latest year.

Figure 3. Trends in shares of public social transfers

168 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD Terms of References and the methodology used by national sources:

A first reason for differences comes from the treatment of negative income. In its own publications Netherlands Statistics do not carry through this treatment. Negative incomes are existent with self employed and also (but rare) with employees and benefit claimants (pay back amounts) whereas in the OECD terms of reference once equivalent household member adjustments are done, all individual components of market income (EH, ES, EO, K, SE) showing negative values should be set to zero.

Difference in the definitions of equivalence scales is also another reason.

Methodological differences between the OECD reference series based from SEP and the others sources:

As student households and households without income throughout the year are excluded from the target population in the Income Panel Survey (IPO), the data report lower levels of income inequality and poverty.

Also an issue related to Additional Enquiry on the Use of (Public) Services used by LIS relies on the underrepresentation of low income groups which lead to underestimation of income disparities and inequality.

5. Summary evaluation

The Dutch Socio-Economic Panel (SEP) seems to be the most appropriate source for income distribution and poverty data in the Netherlands due to the quality of the data and the scope of the reference population.

However in 2006 there is a break in series due to a policy change in the system of Health cost insurance (excluding employers’ contributions to social security) which makes figures not comparable over time.

Bibliography

LIS Documentation: Social and Cultural Planning Office (SCP): Cross national data center in Luxembourg http://www.lisdatacenter.org/.

The Netherlands in a European Perspective Social & Cultural Report 2000.

The Poor Side of the Netherlands, Results from the Dutch ‘Poverty Monitor’,,1997-2003,Cok Vrooman (ed.),Stella Hoff (ed.), Social en Cultural Statistics Netherlands, Planning Office, The Hague, June 2004.

University of Amsterdam Amsterdam Institute for Advanced labour studies income Distribution dynamics in the Netherlands in the 20th Century Long run developments and cyclical propertires. Emiel Afman, Amsterdam Institute for Advanced Labour Studies, University of Amsterdam, Amsterdam Working Papers Number 05/38.

Statistics Netherlands, Annual Report for 2011, 23 March 2012.

169 INCOME DISTRIBUTION DATA REVIEW – NEW ZEALAND

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for New Zealand are provided by Statistics New Zealand, based on the Household Economic Survey (HES). Since 2001 surveys relates to the year ending 30 June. Previously surveys are for the year ended 31 March. Income reported in survey is the actual amount received in the 12 months before the interview date. In the HES households are interviewed over a period of 12 months (i.e. for 2009-10 from the 1 July 2009 to the 30 June 2010). The income collected therefore covers a 2 year period (i.e. for 2009-10 from 1 July 2008 to 30 June 2010). The OECD income distribution database contains data for the following years: 1985-86, 1990-91, 1995-96, 2000-01, 2003-04, 2008-09 and 2009-10.

1.2. National reporting and reporting in other international agencies:

Income distribution and poverty indicators are also available in the report Household Incomes in New Zealand: Trends in Indicators of Inequality and Hardship 1982 to 2011 by the New Zealand Ministry of Social Development (MSD). (http://www.msd.govt.nz/about-msd-and-our-work/publications- resources/monitoring/household-incomes/index.html) The report uses the same data source as above, i.e., the Household Economic Survey.

While the HES is the generally used source for calculation of poverty rates and inequality indicators in New Zealand, there is also Survey of Family, Income, and Employment (SoFIE). This is a longitudinal survey with sample members interviewed once a year over an eight year period from 2002 to 2010, which also collected information on income and wages. Data from the first seven waves was used for Dynamics of Income and Deprivation in New Zealand, 2002‐2009 conducted within the Health Inequalities Research Programme, University of Otago.

Another survey used to collect information on income in New Zealand is the New Zealand Income Survey (NZIS), which is run as a supplement to the Household Labour Force Survey each year in the April to June quarter. The majority of published data from this survey relates to individuals, although there is one table on household income. However, the NZIS reports on 'weekly income' and relates specifically to an average week during the June quarter; that is a snapshot in time. According to Statistics New Zealand conversion of this weekly income into an annual equivalent is not recommended and the Household Economic Survey provides a better source of annual income. Hence the data are not used in the comparisons in the following sections.

The main characteristics of the surveys are shown in the table below:

170 Table 1. Characteristics of datasets, New Zealand

National surveys (Income)

Household Economic Survey of Family, Income, New Zealand Income Name Survey (HES) and Employment (SoFIE) Survey (NZIS)

Name of the responsible Statistics New Zealand Statistics New Zealand Statistics New Zealand agency Annually from 1973 to Year (survey and 1998, then triennially to Eight annual waves from Each year since 1997 in income) 2007, annually again 2002 to 2010 the April to June quarter since then. Income for the 12 months Income for the 12 months prior to the interview date. prior to the interview date. Period over which Weekly income for an Interviews are carried out Interviews are carried out income is average week in the April over a 1 year period, over a 1 year period, hence assessed to June quarter hence income covers a income covers a two year two year period. period. Income data is collected Covered Income data is collected for all individuals aged population for all individuals aged 15+. 15+. 3126 households Initial sample of c.11 500 Sample size (achieved after c. 15 000 households responding households imputation)

Sample Cross-sectional survey Longitudinal survey procedure

77% for the initial wave, Response rate 68.8% after imputation with an attrition rate of 63% by wave seven

Imputation of Introduced in 2009/10 and missing values applied back to 2006/07

Unit for data Individual 15+ Individual 15+ Individual 15+ collection

Change in the selection of the head of the household since 2008/09 - Break in series interpolation possible 2008/09 data are available in old and new definition http://www.stats.govt.nz/surveys http://www.stats.govt.nz/brows _and_methods/completing-a- http://www.stats.govt.nz/surveys e_for_stats/people_and_comm survey/faqs-about-our- _and_methods/our- Web source: unities/Households/household- surveys/survey-of-family- surveys/nzis-resource.aspx economic-survey.aspx income-and-imployment.aspx

171 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients

Both the OECD reference series and the MSD publication are based on the HES data. The Gini coefficients of the OECD reference series are slightly higher than those shown in the MSD’s publication, although the overall trend is similar. According to Bryan Perry, one of the authors of the MSD publication, a small difference arises because the OECD uses a different equivalence scale, the impact of which this is to raise the Gini by around 0.008 each year. However, it is worth noting that the difference is somewhat greater than 0.008 in earlier years and it appears that the two series are converging. Unfortunately, it is not possible to find a time series of Gini coefficients based on the SoFIE data.

Figure 1. Gini coefficients, New Zealand

OECD Income Distribution Database National series

0.40

0.35

0.30

0.25 1990 1995 2000 2005 2010

2.1.2 Time series of poverty rates, poverty composition

The OECD reference series of income poverty rates, at both the 50% and 60% threshold, are also consistently slightly higher than those shown in the MSD’s publication. It seems likely that this could again be, at least partially, explained by the use of a different equivalence scale. There is also some variation in how close the two series are at different points in time, especially for the 50% threshold. Some of this may be explained by the fact that the national series is rounded to the nearest whole percent, where as the OECD series is correct to one decimal place.

Poverty rates from the SoFIE data were published by the Health Inequalities Research Programme, University of Otago in Dynamics of Income and Deprivation in New Zealand, 2002‐2009. However, these are based on gross equivalised household income, that is, household income from all sources before the deduction of taxes but including all reported transfers, adjusted for household size and composition. This makes it difficult to compare them with the OECD series. As can be seen from the two graphs below, the SoFIE series sits somewhere between the OECD series based on disposable income and that based on

172 market income. However, at the 50% threshold it is closer to the disposable income series whereas at the 60% threshold it is closer to the market income series, even coinciding in 2008-09.

Figure 2. Poverty rates (50% threshold), New Zealand

Figure 3. Poverty rates (60% threshold), New Zealand

The comparison of child poverty rates (50% threshold) in the national series with those of the OECD reference series shows a similar picture to that of the Gini coefficients. As can be seen in the graph below, the OECD series sits above the national series, but appears to converge towards the end of the series.

173 Figure 4. Poverty rates 0-17 years (50% threshold), New Zealand

OECD Income Distribution Database - disposable income National Series 16.0%

14.0%

12.0%

10.0%

8.0%

6.0%

4.0%

2.0%

0.0% 1990 1995 2000 2005 2010 In the past, MSD’s publication has shown poverty rates using only a 60% threshold. This has led to apparent inconsistencies when referring to poverty rates in New Zealand, particularly for the 65+ age group. Historically, the flat-rate pension in New Zealand was above the 50% median threshold so old-age poverty rates using the 50%-threshold got close to zero but were much higher under the 60%- median threshold. Further, the rising New Zealand median through to 2008/09 led to the universal NZ Superannuation falling from above to just below the 50% threshold. This gave rise to a very sharp increase in poverty rates for that year, in both the national series and the OECD series, as can been seen below.

Figure 5. Poverty rates 65+ (50% threshold), New Zealand

MSD’s current year publication shows a table of poverty rates for 65+ at the 50% threshold using both the national methodology and the OECD methodology. The figures for OECD methodology do not match those above, notably for 2004 – 1.5% vs. 9%. The table and text, from page 136 of the report, are shown below:

Table 2 shows the proportion of older New Zealanders (65+) in households with incomes under two commonly used ‘poverty lines’. The top line uses the OECD equivalence scale to ensure consistency with OECD publications. The second line uses the same 50% of median threshold but the Revised Jensen scale as in the rest of the report.

174 Table 25. Proportion of older New Zealanders (65+) in households with BHC incomes below low-income thresholds (‘poverty lines’), set at 50% and 60% of the median in the survey year (%)

1984 86 88 90 92 94 96 98 01 04 07 08 09 10 11

50% OECD equiv 2 2 8 2 1 1 1 3 2 9 18 16 22 13 11 50% NZ equiv 1 1 1 1 1 1 1 2 2 3 8 5 14 5 4 60% NZ equiv 14 17 25 20 3 1 3 25 20 37 38 39 37 36 33

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average incomeK SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2009 natcur 20,066 9,079 4,141 6,359 5,182 6419 (10,580) 40,665 % av HDI 49% 0% 16% 13% 16% -26%

Figure 44. Figure 6 Trends in shares of public social transfers

4 % transfers in disposable income (OECD reference series) % cash social spending in Net National Income (OECD SOCX) 2

0 1990 1995 2000 2005 2010

175 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Differences between OECD series and national series reported by MSD:

The main difference between the two data sets is that the equivalence scale used in the analysis in the MSD paper is the 1988 Revised Jensen Scale, which is very close to what has come to be known as ‘the modified OECD scale’. A single person unit has an equivalence scale value of 1.0. A household of a couple and no children is rated at 1.54, meaning that such a household is considered to have 1.54 equivalent adults. A two adult, two children household is rated as 2.17.

Differences between OECD series and series based on SoFIE:

Methodological differences between OECD reference series based on HES and results from SoFIE are many and include:

 Results from SoFIE were not weighted to the New Zealand population and relate only to the SoFIE survey balanced panel sample.

 SoFIE poverty rates based on gross equivalised household income

 Household income from SoFIE was equivalised using the 1988 Revised Jensen Scale

 Most analyses in SoFIE – unless otherwise noted – used (nominal) equivalised household income calculated before housing costs and did not adjust for changes in Consumer Price Index (CPI).

5. Summary evaluation

The differences between the MSD data and the OECD data are generally quite minor and can be mostly explained by the use of a different equivalence scale for income.

Concerning poverty rates for 65+ the difference between the rates published in the MSD paper as ‘OECD’ 50% threshold, i.e. using OECD definitions, and the rates that we have in the database is quite significant especially for 2004. This should be investigated further.

The methodology used for the SoFIE analysis is too different for the results to be directly comparable with OECD benchmark series. Further, the authors of the University of Otago paper from which the SoFIE poverty rates are taken, point out that they are not arguing that the SoFIE data provides the best evidence on current trends in poverty but rather that it can be used to provide a longitudinal study of dynamics to complement cross-sectional studies using HES data.

176 INCOME DISTRIBUTION DATA REVIEW – NORWAY

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD Income Distribution Data for Norway are computed by Statistics Norway and are based on the Income Distribution Survey.

In the OECD database, income inequality and poverty rates are currently available for the following years: 1986, 1995, 2000, 2004, 2008, 2009 and 2010. No breaks in series occur.

1.2. National reporting and reporting in other international agencies:

 EUROSTAT has been computing indicators on inequalities and poverty for Norway from 2003 (income year 2002) onwards.

 Norway has been included in the EU-SILC (Statistics on Income and Living Conditions) survey since 2003 onwards (income year 2002). EU-SILC is a multi-dimensional instrument focused on the income and the living conditions of different types of households. It is collecting, on an annual basis, timely and comparable multidimensional micro-data on income, material deprivation, housing condition, labour, education, health and subjective well-being.

 The Luxembourg Income Study Database (LIS) includes Norway for the years 1979, 1986, 1991, 1995, 2000 and 2004. It is based on the Microcensus survey that is presented in more details in the below table.

 Statistics Norway is the official national survey in Norway and has been conducting the Income Distribution Survey annually from 1986 to 2004 based on a representative sample survey collected from various Living Condition Surveys and Household Budget Surveys. Since 2005, Statistics Norway produces a totally register-based household income statistics.

The below table presents the main characteristics of those four datasets:

177 Table 26. Characteristics of datasets used for income reporting, Norway

OECD reference series LIS database Statistics Norway Eurostat income distribution database Name Income Distribution Income Distribution Income statistics for EU-SILC Survey Survey households Name of the Statistics Norway; Statistics Norway; Division for Income and Eurostat responsible Division for Income and Division for Income and Wage Statistics agency Wage Statistics Wage Statistics Year (survey and 1990-2010 (data missing 1986, 1991, 1995, 2000, 1986-2010 2003-2010 survey income/wage) for 1991, 1993, 1997, 2004 years representing 1999, 2001, 2003 income for years 2002- income years) 2009. Period over which Annual income During second quarter Annual income Annual income N-1 income is two years after the assessed income year of 2004. Covered All persons residing in All persons residing in population Norway and resident in Norway and resident in private households as of private households as of 31st December of the 31st December of the fiscal year. fiscal year. Sample size total resident population 13131 households Achieved sample size: per 31 December containing 33989 5227 households (4.7mil.) individuals. Sample procedure Cross-sectional Cross-sectional systematic one-stage random sampling design Response rate 43.83% Imputation of No missing income data Not applicable missing values Unit for data Household Individual Individual Individual aged 16+ collection

Break in series No No No No

Web source: http://www.oecd.org/els/ http://www.lisdatacenter. http://www.ssb.no/iffor_en/ http://epp.eurostat.ec.e socialpoliciesanddata/in org/wp- uropa.eu/portal/page/po comedistributionandpov content/uploads/our-lis- rtal/income_social_inclu ertydatafiguresmethodsa documentation-by-no04- sion_living_conditions/q ndconcepts.htm survey.pdf uality/national_quality_r

eports

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below figure shows the evolution of Gini coefficients for Norway from 1990 to 2010, as reported by the OECD, LIS, Statistics Norway and the EU-SILC.

178 Figure 45.1 Trends in Gini coefficient (disposable income)

OECD reference series (IDS) LIS (IDS) Statistics Norway (IDS) Eurostat (EU-SILC) OECD reference series adjusted 0.34

0.32

0.30

0.28

0.26

0.24 Gini (at disposable income) (at disposable Gini 0.22

0.20 1990 1995 2000 2005 2010

According to the OECD reference series, income inequality in Norway rose steadily from 1990 to 2004 from 0.228 to 0.280. Year 2005 then witnessed a single large increase by some 4 points, reaching 0.326, before a further single impressive decline to 0.240 in 2006, thus reaching levels similar to the 1990s.

The fluctuations in the years 2004-2006 can be explained by an increase between 2002-2005 in dividends at the top of the income ladder, which were then corrected in 2006 by the introduction of higher taxes on dividends. This is related to tax reform.

During the first half of the 2000s income from dividends increased enormously in Norway. If looking at the macro amounts received by households (in billion NOK) they increased from 13,2 in 2001 to 99,4 in 2005, to fall to 7,4 in 2006. Since dividends are extremely unequally distributed (90% is received by the top 2% of households), taxes on this kind of income were increased - starting from income year 2006. (Dividends were tax-free income for the share holders 2002-2005, but the companies were taxed on their profit). From 2006 dividends are taxable income for the individual share holder as well. The tax reform of 2006 was announced well in advance so that the companies and share holders had plenty of time to adjust (i.e. they more or less paid out as much dividends as they could in the years 2002-05, and almost nothing in 2006). Things are gradually getting back to normal in the years after 2006 (i.e. 25 billion NOK were received in dividends in 2008) (information received by Jon Epland, Statistics Norway).

To adjust for these movements, Statistics Norway provided an “adjusted” series of Gini coefficients to the OECD which is based on top coding. As can be seen, this resulted in a smaller shock to the Gini coefficients although the hook can still be seen.

The other series show similar levels and trends as the OECD reference series. The series from Statistics Norway is identical to the OECD series as they both use the same sources and equivalence scale. The LIS series and EU-SILC series show lower levels of income inequality, but the trends remain quite similar.

Also, when comparing the income quintile share ratio (S80/S20) from the OECD series with the series from the EU-SILC and Statistics Norway, it is visible that there is much more fluctuation in the levels and trends of the latter two, particularly marked during the 2004-2006 period. The OECD series is overall quite

179 steady, rising slightly over the whole period from 3.5 points in 1995 to 3.7 points in 2010. The trends of the OECD series are similar to the Statistics Norway series over the last three year.

Figure 1.2 S80/20

Looking at the P90/P10 Index, data is available for the OECD, Statistics Norway and LIS. The three series show similar trends between 1990 and 2010, all ranging between 2.5 and 3 points, with only minor variations. However, the OECD reference series is consistently higher than the Statistics Norway series, reaching 2.97 points in 2008, as opposed to 2.8 points for the Statistics Norway series.

Figure 1.3 P90/P10

180 2.1.2 Time series of poverty rates

The OECD data between 1990 and 2004 can be compared only with the LIS series. There are some discrepancies between 1992 and 1998 but not thereafter.

According to the OECD income distribution database, the share of the Norwegian population living with less than 50% of the median equivalised income (163.711 Kroners per year in 2008) has increased from 6.8% in 2004 to 7.8% in 2008, before dropping to 7.5% in 2010.

The EU-SILC series contrasts with the OECD series for the early 2000s but not thereafter. The Statistics Norway series, while lower than the OECD series, shows an increase in poverty rates between 2004 (6.5%) and 2008 (7%) before declining again in 2009 to 6.6%.

Figure 2.1 Trends in poverty rates

OECD reference series (IDS) LIS (IDS) Statistics Norway (IDS) Eurostat (EU-SILC) 10.0%

8.0%

6.0%

4.0%

2.0% Poverty rate (at 50% median income) median (at 50% rate Poverty

0.0% 1990 1995 2000 2005 2010

As for child poverty, the series show different levels with some exceptions (about 1 to 2 points), although all data seem to indicate a trend increase over the last decade. Indeed, the OECD series increased from 3.6% in 2000 to 5.1% in 2010. The EU-SILC series increased from 4.2% in 2002 to 6.6% in 2009. The LIS series increased from 3.6% in 2000 to 5.3% in 2004. The Statistics Norway series shows lower levels of child poverty rates, with 4.6% in 2010, following a decline since 2008 (5%).

181 Figure 2.2 Trends in Child poverty rates

OECD reference series (IDS) LIS (IDS) Statistics Norway (IDS) Eurostat (EU-SILC) 10.0%

8.0%

6.0%

4.0%

2.0% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 0.0% 1990 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Except for the share of capital income, the shares of income components match well with comparable calculations from EU-SILC.

Table 27. Shares of income components in total disposable income, OECD reference series

Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period, except for the period between 2004 and 2008 where SOCX data suggest a decline, in contrast to the income micro data.

182 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and the other income series:

5. Summary evaluation

Broadly speaking, the different data sources of indicators for Norway follow similar trends throughout the covered period. OECD and Statistics Norway data generally match. The exception being poverty rates, where the OECD series show a different trend to LIS data in the early 1990s and a different trend to Eurostat data in the early 2000s. Yet, there is convergence of series in the last two years, thus closing the gap with the other series.

The minor remaining discrepancies between the different series can be explained by the different methodology, with the OECD and LIS series using a square root equivalence scale, and the EU-SILC series and Statistics Norway using an OECD modified equivalence scale.

183 INCOME DISTRIBUTION DATA REVIEW – POLAND

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Poland are based on the Eurostat Survey of Income and Living Conditions (EU – SILC) since income year 2004 and currently going up to 2009. Earlier data for 2000 are also included in the database and are based on the Household Budget Survey. The results from these two data series are considered not comparable and, hence, no trend comparison between 2000 and later years are published by OECD.

1.2. National reporting and reporting in other international agencies:

The Household Budget Survey (HBS) from the CSO, Poland provides information for the analysis of the living condition of the population. This survey has been the source for several studies on income inequality. A comparison of different results obtained from the HBS is proposed in Brzezinski and Kostro (2010). This paper proposes in particular a comparison of the evolution of the Gini coefficient for Poland according to various sources, which are actually various ways of using the HBS data as the main data set: some studies use individual data while others use grouped data or a mix of both; and different equivalence scales are used across the study. In summary:

 Keane and Prasad (2002) use individual data and apply food-share based equivalence scale

 Milanovic (1999) uses both grouped and individual data and applies simple per capita scale

 UNICEF (2009) uses grouped data and per capita scale.

 Daras et al. (2006) uses individual data and OECD equivalence scale.

 Szulc (2000; 2003) uses individual data and an empirically estimated scale

Furthermore, the Luxembourg Income Study (LIS) also uses HBS as the basis of their standardised income micro database (currently available for 1986, 1992, 1995, 1999 and 2004).

The key characteristics of the two main surveys used for reporting on household incomes in Poland, HBS and EU-SILC are shown in the table below:

184 Table 28. Characteristics of datasets, Poland

Household Budget Survey (HBS) Eurostat Survey of Income and Living Conditions (EU – SILC)

Name Household Budget Survey (HBS) (Badania EU-SILC Budzetow Gospodarstw Domowych) Name of the Household budget survey is conducted by Eurostat responsible statistical offices. However, the responsibility for agency the survey content and coordination lies with the Central Statistical Office, Social Surveys and Living Conditions Statistics Department in cooperation with Statistic Office in Łódź which specializes in living condition statistic. Year (survey Since 1986 (at least) Survey years: 2005-2011, referring to and incomes in 2004-2010. income/wage) Period over All through the year. Every month of the year, a May-June which income different group of households participates in the is assessed survey. Covered HBS covers 5 socio-economic groups of the The reference population of EU-SILC is all population population: employees' households, farmers' private households and their current households, households of the self-employed, members households of retirees and pensioners, residing in the territory at the time of data households living on unearned sources. The collection. Persons living in collective households of foreign citizens with permanent or households and in institutions are long-lasting residence in Poland and using Polish generally excluded from the target language participate in the survey. Households population. living in collective homes, like students' hostels, social welfare homes as well as households of the diplomatic corps of foreign countries are not covered. Sample size In 2011 there were 3132 dwellings surveyed every 16250 month and thus it was planned to achieve the results for the whole year from households inhabiting 37 584 dwellings. Actually number of surveyed households was 37 375 Sample Two-stage stratified sampling scheme. The Stratified multi-stage sampling. The procedure sampling units for the first stage is the area survey primary sample units are the enumeration points (as designed for the National Census) and census areas. The secondary sample units those for the second stage are dwellings. are dwellings. All the households from the selected dwellings are supposed to enter the survey. Households are clusters of individuals and all members aged 16 and over at the end of the income reference period of a selected household are eligible for inclusion in the sample. The source of the sampling frame is the Domestic Territorial Division Register (TERYT system). 4-year rotational integrated design. Response rate About 55% in 2011 Household response rate: 63.5% Individual response rate: 62.28 % Imputation of Selected households not participating in the missing values survey are replaced by randomly selected other households. Unit for data Households Households and households’ members. collection

185 Household Budget Survey (HBS) Eurostat Survey of Income and Living Conditions (EU – SILC)

Break in series - 1993: HBS became fully representative for all main socio-economic types of households and a new method of rotating households was applied: monthly rotation replaced previously used quarterly rotation. - 1997 (experimentally) and 1998 (definitively): New definitions of some core concepts (i.e. disposable income) were implemented in order to adjust HBS to Eurostat recommendations. [Source: Brzezinski, M. and K. Kostro, Income and consumption inequality in Poland, 1998-2008, University of Warsaw, Faculty of Economic Sciences, Bank i Kredyt 41 (4), 2010, 45-72] Web source: http://www.stat.gov.pl/cps/rde/xbcr/gus/LC_house http://epp.eurostat.ec.europa.eu/portal/pag

hold_budget_survey_in_2011.pdf e/portal/income_social_inclusion_living_co

nditions/quality/national_quality_reports

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Figure 1.1 compares series of Gini coefficients based on EU-SILC with a selection of series based on the HBS. The two series based on EU-SILC (OECD and Eurostat) are quasi identical despite the use of slightly different equivalence scales (square root scale for OECD series, and modified OECD-scale for Eurostat series).

As mentioned in the first section, the literature proposes several alternative Gini coefficient series based on the HBS. For various reasons, however, none of them can be considered as fully comparable for comparison with EU-SILC figures, because of different data source patterns (e.g. the period over which income is assessed) but also because of differences in methodology (e.g. adjustment via different equivalence scales).

Because of their rather long-period availability, UNICEF data (based on HBS) have been included in the chart below, but the comparison should be conducted with caution. Furthermore, data prior to 1997 are not comparable with more recent data because of changes in HBS concepts. UNICEF data rely on per- capita incomes.

Gini coefficients from Daras et al. And from Brzenski and Kostro (both drawn from HBS) are based on the OECD equivalence scale, which is closer to the square root scale used by OECD than the simple per capita scale. Further, Gini coefficients from LIS are based on the square root scale. These coefficients would, thus, have been better candidates for a comparison exercise. However, if we exclude data prior to the HBS methodological break, they are available in the paper only for a few data points and, with the exception of Brzenski and Kostro only up to 2004 and do not allow to capture accurately the movements in the income distribution.

186 All these four HBS-based series suggest a signficiant increase in income inequality between the late 1990s and 2004. Nevertheless, the level of the Gini coefficient from UNICEF is significantly higher than those from the other three series.

From 2003, also HBS-based series published by the Polish CSO are available. This series uses the same equivalent scale as Eurostat (modified OECD-scale). In 2004, which is the only year for which almost all the series are available, EU-SILC data are roughly in the middle of the gap between the HBS- based series from UNICEF and those from LIS and Daras et al, and are very close to the level suggested by official CSO figures.

After 2004, EU-SILC based data show a steep decline in Ginis, lasting until 2008 at least. This is somewhat in contrast to all HBS-based series which show a much more modest fall or stability during this period. The official CSO series and that published by Brzenski and Kostro show a slight decrease which remains below 0.01 point and is basically concentrated in 2006. Only the UNICEF series suggest a larger decrease (about 0.02 points), again predominantly taking place in 2006. Given that EU-SILC series suggest the main bulk of the decrease in 2005 but similar trends for after that year, the picture below suggests that there may be an issue with the EU-SILC data in the first year of implementation, 2004, resulting in some overestimation of the level of inequality.

Two other inequality indicators are charted below, the S80/S20 and P90/P10 ratios, respectively in Figures 1.2 and 1.3. They both confirm the picture drawn above.

Figure 1.1 Gini coefficients, Poland

OECD reference series (based on SILC) Eurostat (based on SILC) CSO (based on HBS) LIS (based on HBS) UNICEF (based on HBS) Daras et al. (based on HBS) Brzezinski and Kostro (based on HBS)

0.40

0.35

0.30 Gini (at disposable income) (at disposable Gini

0.25 1990 1995 2000 2005 2010

187 Figure 1.2 S80/S20, Poland

OECD reference series (based on SILC) Eurostat (based on SILC) CSO (based on HBS)

7.0

6.5

6.0

S80/S20 5.5

5.0

4.5 2000 2002 2004 2006 2008 2010

Figure 1.3 P90/P10, Poland

OECD reference series (based on SILC) Eurostat (based on SILC)

LIS (based on HBS) 5.5

5.0

4.5

4.0

3.5 P90/P10

3.0

2.5

2.0 1992 1994 1996 1998 2000 2002 2004 2006 2008 2010

2.1.2 Time series of poverty rates

Poverty rates measured by the CSO are based on several poverty lines, including:

 The relative poverty line which is equal to 50% of mean equivalised expenditure.

 The legal poverty line

 The extreme poverty line

From these three, the closest concept to what is commonly available from the OECD income distribution database and the EU-SILC is the relative poverty line. Only this series was kept for comparison with other sources although, for the CSO series, the threshold is proportionate to equivalised expenditure rather than equivalised disposable income (after social transfers).

188 Because of the difference in concept, the poverty rate compiled by the CSO cannot be compared to the EU-SILC indicator in terms of absolute level. Comparison in terms of annual changes is more relevant. Both poverty rates show a significant decrease between 2004 and 2007, more pronounced in EU-SILC data (-4.3%) than in HBS data (-3 %). For the 50%-median threshold, this decrease is followed by a period of relative stability and, in the case of the OECD series, a slight increase in 2008. For the 60%-median threshold, EU-SILC based series suggest a slight increase in 2008 and 2009.

Figure 2.1 Poverty rate (50%) Poland

OECD reference series (based on SILC) Eurostat (based on SILC)

CSO (based on HBS) LIS (based on HBS)

22.0%

20.0%

18.0%

16.0%

14.0%

12.0%

10.0% Poverty rate (at 50% median income) median (at 50% rate Poverty 8.0% 1999 2001 2003 2005 2007 2009 2011

Figure 2.2 Poverty rate (60%) Poland

OECD reference series (based on SILC) Eurostat (based on SILC) LIS (based on HBS)

22.0%

21.0%

20.0%

19.0%

18.0%

17.0%

16.0%

15.0% Poverty rate (at 50% median income) median (at 50% rate Poverty

14.0% 1999 2001 2003 2005 2007 2009 2011

2.2 Wages

See Part II of the present Quality Review.

189 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. Unfortunately, such information could not be found for the HBS data source described in table 1.

Table 2. Shares of income components in total disposable income, OECD reference series

Year Unit Wages Capital Self Employment Transfers Taxes Disposable income (HDI) 2008 natcur 20,815 393 3,308 6,835 -7,176 24,114 % av HDI 86% 2% 14% 28% -30%

Figure 3 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. There are significant trend differences between both series during the period 2005 through2008: during those years macro estimates of transfer shares (SOCX) increased while micro estimates (EU-SILC) decreased considerably.

Figure 3 Trends in shares of public social transfers

10 % transfers in disposable income (OECD reference series) 5 % cash social spending in Net National Income (OECD SOCX)

0 1990 1995 2000 2005 2010

4. Metadata of data sources which could explain differences and inconsistencies

Survey characteristics differ between HBS and EU-SILC. While response rates are rather similar (though slightly higher for EU-SILC), HBS has a larger sample size and, in particular, assesses household incomes on a monthly basis. In HBS, data are collected all through the year for different groups of households every month for a total of 37375 households surveyed in 2011, while 16250 households are surveyed in EU-SILC, with incomes assessed over a 2-month period (May-June) but for the total period of the preceding year.

190 Comparing the indicators compiled by various sources on the basis of these two surveys, differences between EU-SILC and various HBS-based sources can also be explained by significant differences in data treatment, especially in the equivalence scales, or the treatment of extreme income values (including negative incomes).

5. Summary evaluation

HBS and EU-SILC (since 2004) are the two main data sources for monitoring household incomes in Poland. Different series calculated from data from HBS show the same increasing trend of inequality for the period between the late 1990s and 2004. For the period between 2004 and 2008, EU-SILC data suggest a significant decline and different HBS data suggest a much more modest decline in inequality, including the official CSO series. For the latest years 2009 and 2010, all data sources indicate stability in the development of inequality. Given that EU-SILC series suggest a main inequality decline particularly from 2004 to 2005 but roughly similar trends for after that year, there may be an issue with the EU-SILC data in the first year of implementation, resulting in some overestimation of the level of inequality. Furthermore, the official CSO series suggests a significantly higher level of inequality (3-4 points higher Ginis, 1-1.5 point higher S80/S20 ratios – these are significant discrepancies)

HBS presents the advantage of being available over a relatively longer span of years but with methodological or conceptual breaks in 1993 and 1997, while the number of available year from EU-SILC (from 2004) is still limited. On the other hand, the response rate is higher for EU-SILC (more than 60%) than for HBS (about 55%) and it possesses a set of variables which is needed for detailed income analysis (e.g. standardised components of income).

Bibliography:

Brzezinski, M. and Kostro (2010), “Income and consumption inequality in Poland, 1998-2008”, Bank i Kredyt 41 (4), 2010, 45-72. http://www.bankikredyt.nbp.pl/content/2010/04/bik_04_2010_03_art.pdf

Keane M.P., Prasad E.S. (2002), Inequality, Transfers and Growth: New Evidence from the Economic Transition in Poland, Review of Economics and Statistics, 84 (2), 324–341.

Milanović B. (1999), Explaining the Increase in Inequality during Transition, Economics of Transition, 7 (2), 299–341.

UNICEF (2009), TransMONEE 2009 Database, UNICEF Regional Office for CEE/CIS, Geneva.

Daras T., Zienkowski L., Żółkiewski Z. (2006), Zróżnicowanie dochodów i sfera ubóstwa w Polsce w latach 1993–2004, Bank i Kredyt, 37 (11–12), 27–45.

Szulc A. (2000), Economic Transition, Poverty, and Inequality: Poland in the 1990s, Statistics in Transition, 4 (6), 997–1017.

Szulc A. (2003), Inequality During the Economic Transition in Poland: 1993–1999 Evidence, unpublished anuscript

191 INCOME DISTRIBUTION DATA REVIEW – PORTUGAL

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income data currently available for Portugal refer to income earned in the years 1979-80, 1990, 1995, 2000, 2004, 2005, 2006, 2007, 2008 and 2009. Over time, the OECD used two different types of data sources for reporting on income inequality and poverty in that country.

 From 1979 to 2004, OECD estimates were based on data from Household Budget Survey conducted by the National Statistical Office (Instituto Nacional de Estatística - INE).

 From 2004 onwards (income year: 2003), OECD estimates have been based on the EU Survey on Income and Living Conditions (EU-SILC).

There is a strict break in series due to the change in source to EU-SILC in 2004, with no overlapping year with two data sources available. Data prior to 2004 are strictly not comparable with data from the 2004 onwards. Note that this break is shown in the Figures below between 2000 and 2004.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

The following datasets are available for Portugal:

 The National Statistical Office has been computing the Household Expenditure Survey, labelled Inquérito às Despesas das Famílias in Portuguese, since 1960s. On the INE’s website, only the two latest publications are available on-line in Portuguese language (2005/2006 and 2010/2011). The latest available publication is accessible on the following url: http://www.ine.pt/xportal/xmain?xpid=INE&xpgid=ine_publicacoes&PUBLICACOESpub_boui =141619793&PUBLICACOEStema=00&PUBLICACOESmodo=2

 The National Statistical Office has also been computing the Survey on Income and Living Conditions called Inquérito às Condições de Vida e Rendimento since 2004 which is presenting EU-SILC data. Data are available on-line and present the different indicators estimated at the European level. Access is available on the following url: http://www.ine.pt/xportal/xmain?xpid=INE&xpgid=ine_indicadores&indOcorrCod=0004206&c ontexto=bd&selTab=tab2

1.2.2 International reporting:

 Eurostat has been calculating on an annual basis indicators on inequalities and poverty for Portugal since 1995 (income year: 1994) onwards, based on the European Community Household Panel (ECHP). From 2004 (income year: 2003) onwards, Portugal has been included in the EU- SILC (Statistics on Income and Living Conditions) survey. The EU-SILC survey aims at studying poverty, social exclusion and living conditions on the basis of indicators that can be compared at the European level.

192 Two specificities in terms of definition and covered population are specified in the EU-SILC methodology for Portugal22:

 The following specificity for “Household membership” is taken into account in Portugal: “Contrary to the EU-SILC concept, persons absent for long periods, but having household ties (persons working away from home) are not considered as household members if the absence is for more than 6 months (the income obtained from them is considered as a private transfer).

 Collective households and institutions are excluded from the covered population (represent 1% approx. of the total population residing in Portugal).

Please, note that Portugal is not a member of the Luxembourg Income Survey database.

The below table 1 presents the main characteristics of those datasets:

Table 29. Characteristics of datasets used for income reporting, Portugal

Name Household Expenditure Survey - Inquérito às Condições Inquérito às Despesas das EU-SILC from Eurostat de Vida e Rendimento Famílias Name of the National Statistics Office (Instituto Eurostat INE and Eurostat responsible Nacional de Estatística) agency

Year Income years: 1999, 2005 and Annually since 2004 (income year : 2003) Same than for EU-SILC 2009 Data Between March 1st 2010 Between April and November. Same than for EU-SILC collecting and February 27th 2011 (in Região Autónoma da Madeira it began final March 2010) (wave 2010 – 2011);

Covered Unknown The Master Sample for Portugal is designed Same than for EU-SILC population on the Census of Population and Housing (Census/2001). It is constituted by almost 750000 private dwellings. Collective households and institutions are excluded (represent 1% approx. of the total population residing in Portugal).

Sample size 16,815 (wave 2010 – 2011) Achieved household sample size: 853 (year Same than for EU-SILC 2006); 1951 (year 2007) and 3208 (years 2008-2009). Achieved individual sample size: 2268 (year 2006), 5137 (year 2007) and 8423 (years 2008-2009). Method of Interviews (tbc) Data collection is based on PAPI (Paper Same than for EU-SILC data Assisted Personal Interview) and CAPI collection (Computer Assisted Personal Interview).

22 Eurostat, 2012, “2010 COMPARATIVE EU INTERMEDIATE QUALITY REPORT”, Version 3 – October 2012

193 Overtime, the CAPI mode of data collection has increased up to 96% versus 3.77 for PAPI (wave 2009). Sampling Representative stratified clustered Stratified two-stage clustered sampling Same than for EU-SILC method sample of private dwellings with design. The primary sample unit (PSU) is main residence in the national the area of the Master sample (made of territory, census enumeration areas).The secondary sample unit (SSU) is dwellings. All the households and therefore all the persons living in the same dwelling are interviewed. Sampling Dwellings Dwellings / Addresses Same than for EU-SILC unit Imputation Unknown Yes (through regression method only) Same than for EU-SILC of data Response 56% (wave 2010-2011) Unknown Unknown rates (Considering in the denominator only the eligible and contacted dwellings (14,032), the specific response rate rose to 68%). Remark Available only partially in English Annual income of the year prior to the The national statistics survey. Before 2004, Eurostat was office is calculating these compiling statistics for Portugal since 1994. data for the EU-SILC survey. There is therefore a perfect match between the EU-SILC and the INE website concerning income inequality and poverty rates indicators.

Websource http://www.ine.pt/xportal/xmain?xp http://epp.eurostat.ec.europa.eu/portal/page/ id=INE&xpgid=ine_publicacoes&P portal/income_social_inclusion_living_con UBLICACOESpub_boui=1416197 ditions/introduction 93&PUBLICACOEStema=00&PU BLICACOESmodo=2

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Time series on Gini coefficients are estimated from 1979/80 to 2010 by the OECD. According to the OECD income distribution database, income inequality among total population remained relatively stable in Portugal since 1975. It evolved from 0.35 in 1980 down to 0.33 in 1990 with an increase to the original level by the late 1990s. Since 2004, income inequalities continuously recorded a significant decline, by four points, according to OECD and Eurostat reporting. However, the National Statistical Office published the Gini coefficient for 2010 which recorded a slight increase to 0.34 reversing the previous downward trend. The different time-series on Gini coefficients are presented in Figure 1:

194 Figure 46. Trends in Gini coefficient (disposable income)

OECD references series INE (since EU SILC) - Inquérito às Condições de Vida e Rendimento INE - Inquérito às Despesas das Famílias Eurostat (ECHP) Eurostat (EU-SILC) 0.45

0.40

0.35 Gini (at disposal income) disposal (at Gini

0.30

Please find below some more detailed explanations regarding inequalities and income distribution for each OECD-time series before and after the 2004 break:

 From 1980 to 2000, OECD data cannot be compared with any other time-series as on-line historical series were not available before that date. From 1994 to 2000, OECD data can be compared with Eurostat which compiled Gini coefficients from 1994 to 2000 and, over this period, OECD and Eurostat time-series were matching. The 1999 figure estimated at 0.35 by the National Statistical Office in Inquérito às Despesas das Famílias was also in line with both the OECD and Eurostat time-series.

 From 2004 onwards, OECD data can be compared with EU-SILC data and national statistics: Inquérito às Despesas das Famílias and Inquérito às Condições de Vida e Rendimento. The OECD points overlap both with the EU-SILC and the National Statistical figures coming from the Inquérito às Condições de Vida e Rendimento.

Over this period, inequalities continuously and significantly decreased from 2004 which was characterised by a historical high level of Gini coefficient of 0.38. This downward trend is then shared by the different available datasets as the different lines are showing a perfect similarity. In 2009, Gini coefficients were estimated at 0.33 with a broad consensus between the different time-series. This rate contributed to reflect the improvement in the asymmetry of the income distribution in Portugal over the last years. Therefore, this graph on Gini coefficients is showing a strict convergence of data over time between the different datasets.

The relative steadiness of Gini coefficients up to 2004 and its regular decline from then are also corroborated by the trend of the income quintile share ratio (S80/S20).

The OECD income quintile share ratio was estimated at 5.6 in 1990 and at 5.7 in 2009. In between, the income share ratio recorded up and down moves and it rose up to 6.9 in 2004, the same year that recorded the highest Gini coefficient in Portugal over the time period of reference. From 2004, this ratio has continuously decreased up to now. The latest available data, estimated for 2010, expressed that income

195 of the 20% top income group was 5.7 times the income of the 20% bottom income group, confirming the downward tendency of this indicator.

Figure 47. S80/S20 ratio

OECD reference series Eurostat (ECHP) Eurostat (EU-SILC) INE (since EU SILC) - Inquérito às Condições de Vida e Rendimento INE - Inquérito às Despesas das Famílias 8 7 6 5 4

3 S80/S20 ratio S80/S20 2 1 0

Between 1990 and 2000, S80/S20 ratios slightly increased from 5.6 to 6.2 according to the OECD estimates. Over the same decade, Eurostat calculated also this income quintile share ratio but recorded slightly higher ratios than the OECD: according to the European institution, the ratios moved down from 7.4 in 1994 to 6.5 in 2000. On the contrary, the National Statistics Office stated the 1999 S80/S20 ratio in its publication Inquérito às Despesas das Famílias for 1999 which was evaluated at 5.7, a ratio below the OECD and the Eurostat ones.

Between 2004 and 2010, this OECD income distribution indicator can be compared with EU-SILC and the national survey labelled Inquérito às Despesas das Famílias. Please, recall that the figures presented on the national statistics website (INE website) are also used for EU-SILC such that they are exactly the same. Graphically, the two point lines from EU-SILC and INE (Inquérito às Condições de Vida e Rendimento) are totally merging between 2004 and 2010. The OECD time-series is also perfectly matching with the EU-SILC ones. Therefore, these three data sets are showing a subsequent decline of the income distribution moving from 6.9 in 2004 (OECD estimates) down to 5.7 in 2009 (OECD estimates).

The Inquérito às Despesas das Famílias produced only two points of estimates for S80/S20 ratio in 2005 and 2009. The downward trend is similar than with the EU-SILC and the OECD but in a smaller proportion and the national survey ratio remained below the other datasets. The spread between the Inquérito às Despesas das Famílias and OECD/EU-SILC time-series have seemed to reduce over the last decade underlying a data convergence between the different available reporting for Portugal.

2.1.2 Time series of poverty rates

Time series on relative poverty rates are presented below with a 60% threshold (total population) and with a 50% threshold (total population and children).

According to the 2011 EU-SILC survey, the resident population in risk of poverty was 18.0% in 2010 (60% threshold) which is the latest available data. EU-SILC also presented that “the impact of social

196 transfers (excluding pensions) in the reducing of the risk of poverty has decreased from 8.5 percentage points (p.p.) to 7.3 p. p. in 2010”23 (INE Website - 2012).

You can find below the graph of relative poverty for Portugal with a 60% threshold. OECD time series on this indicator can be compared with Eurostat (before and after the launch of EU-SILC) and with the Inquérito às Despesas das Famílias from the national statistics office. As mentioned previously, the figures presented on the national statistics website (INE Website - Inquérito às Condições de Vida e Rendimento) are exactly the same than for the European survey.

Figure 48. Trends in poverty rates with a 60% threshold

OECD reference series INE - Inquérito às Despesas das Famílias Eurostat (ECHP) INE - Inquérito às Condições de Vida e Rendimento Eurostat (EU-SILC)

24%

22%

20%

18%

16%

14%

12%

10%

With a concentrated scale, we can see graphically that the relative poverty rates with a 60% threshold subsequently slightly decreased over time.

From 1990 to 2000, the OECD time-series can be partially compared with the Eurostat estimates. According to the OECD, the at-risk of poverty rate decreased from 22.29% in 1975 down to 20.56% in 2000. Eurostat demonstrated a similar trend from 1994 with annual estimates: the at-risk of poverty rate decreased from 23% to 20% in 2000. The Eurostat and OECD figures are then perfectly matching over the period.

After 2004, the previous downward trend continued but in a smaller proportion. OECD estimates for relative poverty rates at 60% median decreased from 20.7% (2004) down to 19.2% in 2009 which is the latest available figure in the OECD database. These estimates can be compared with the EU-SILC and with the national study called Inquérito às Despesas das Famílias. As mentioned earlier, the other national study Inquérito às Condições de Vida e Rendimento is totally matching with the EU-SILC estimates and points are overlapping for shared years of reference (2007, 2008, 2009 and 2010).

While comparing with the other databases, the OECD estimates are showing higher poverty rates and a less regular decrease. Indeed, according to the OECD, the at-risk of poverty rates slightly declined from 20.70% in 2004 to 19.20% in 2009 while showing a new peak in 2007 at 20.99%. On the contrary, the

23 INE, 13 July 2012, ”At risk of poverty rate was 18% - 2011”, Press Release in http://www.ine.pt/xportal/xmain?xpid=INE&xpgid=ine_destaques&DESTAQUESdest_boui=132815432& DESTAQUEStema=5414278&DESTAQUESmodo=2

197 European survey (and the Inquérito às Condições de Vida e Rendimento) have been recording the same trend but with lower relative poverty rates over the period: the at-risk of poverty rates indeed declined from 19.40% (2004) down to 18% in 2010 and the 2007 peak is barely readable on the graph (move from 18.10% to 18.50% between 2006 and 2007). Explanations may be based on the differences regarding the terms of references as defined between the two institutions. Indeed, the OECD reference series uses the square root of household size, whereas Eurostat uses the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

With even more lower figures, the national study Inquérito às Despesas das Famílias estimated that the number of households living with less than 60% of the median income moved from 16.40% in 2005 down to 14.80% in 2009.

Figure 4 shows the evolution of the relative poverty rate with a 50% threshold. The evolution of the relative poverty trend with a 50% threshold is quite similar than the one with a 60% threshold.

Figure 49. Trends in poverty rates with a 50% threshold

OECD reference series Eurostat (ECHP) Eurostat (EU-SILC) INE - Inquérito às Condições de Vida e Rendimento 17% 16% 15% 14% 13% 12% 11% 10% 9% 8% 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

The OECD time series can be compared with the Eurostat estimates (before and after establishing EU- SILC). As mentioned previously, the point-lines of the national study Inquérito às Condições de Vida e Rendimento are exactly the same than the EU-SILC ones over the time period of consideration.

From 1975 to 2000, the number of households living with less than 60% of the median income decreased on an elusive manner, from 16.20% in 1975 to 13.67% in 2000. These estimates are matching well with the Eurostat ones.

From 2004 onwards, the OECD is showing a slight decrease of relative poverty rates as the EU SILC and the Inquérito às Condições de Vida e Rendimento from INE did but with fewer steadinesses. According to the OECD, relative poverty rates decreased from 13.18% in 2004 down to 12.01% in 2009 with a same peak in 2007 at 13.61%. In addition, OECD rates appear to be higher than the other datasets but by less than 1 percentage point.

Finally, the below graph dedicated to children poverty in Portugal is partially confirming the two trends described here-before.

198 Figure 50. Trends in Child Poverty rates (50% threshold)

From 1990 to 2000, the relative poverty rates for children increased from 12.37% to 15.57%. This trend cannot be confirmed by other estimates as no other datasets were available over this period of time.

From 2004 onwards, the OECD estimates can be compared with EU-SILC. Both lines are showing similar trends even though the OECD estimates are slightly higher than the EU-SILC ones. The 2007 peak is also relevant for children poverty rate which recorded an increase from 15.12% to 18.74% between 2006 and 2007. Over the time period, the relative poverty rates for children decreased from 17% in 2004 to 15.63% in 2009 (OECD estimates). On average, these rates are higher by 1 or 2 percentage points than the ones that are estimated at the European level.

Please, note that on the Portuguese national statistics website, relative poverty rates for children are only provided at a 60% threshold which is preventing from undertaking a comparison with the OECD data points.

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2009 natcur 6398.31 3204.16 3.88 9,606 372.174 1380.003605 3293.55 -2952.2 11699.87117 % av HDI 54.7% 82.1% 3.2% 11.8% 28.2% -25.2%

199 Figure 6 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show a similar trend throughout the period.

Figure 6 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

No major inconsistancies can be highlighted. However, the following points can be considered with room for improvement for better comparisons:

Differences between the OECD reference series and Eurostat / EU-SILC:

 The OECD time-series has been based on EU-SILC since 2004 (income year). Therefore, unsurprisingly, time-series on gini coefficients, income distribution indicators and poverty rates as published in the EU-SILC are matching well with the OECD since that date;

 That said, OECD and EU-SILC data are presening divergences on the level of relative poverty rates, especially at the 60%-median which seem always lower for EU-SILC than for the OECD (up to 2 points differences). Explanations may be based on the differences regarding the terms of references as defined between the two institutions. Indeed, the OECD reference series uses the square root of household size, whereas Eurostat uses the OECD modified equivalence scale (1.0 to the first adult, 0.5 to the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

 Before the implementation of EU-SILC, Eurostat was calculating income inequality and poverty indicators over the period 1994 – 2003. Although not based on the same source (ECHP for Eurostat, HBS for OECD), data and trends were still pretty similar with the OECD ones.

200 Differences between the OECD reference series and the Household Expenditure Survey (Inquérito às Despesas das Famílias - Instituto Nacional de Estatística):

 The Portuguese National Statistics Office conducts a Household Expenditure Survey which main results are compiled in the publication Inquérito às Despesas das Família which is available on- line. However, this documentation is published only in Portuguese language with the exception of the Executive Summary which is translated in English.

 Data are matching relatively well with the OECD ones while being continuously lower than the OECD ones for the the income quintile share ratios and the relative poverty rates.

 Variations in methodology and terms of references might explain these differences.

Differences between the OECD reference series and the Household Budget Survey (Instituto Nacional de Estatística):

 The OECD time-series on income inequalities and poverty rates were based on the Household Budget Survey from 1980 to 2000. Unfortunately, this survey is not accessible on-line on the National Statistical Office’s website which is prejudicial for undertaking any data comparisions.

5. Summary evaluation

Over the reference period, the OECD time-series are matching pretty well with other national and European reporting. Despite slight differences regarding the figures, general patterns and trends are quite similar. This stems from the integrated way in which the series are produced: the national sources are used to generate EU-SILC, which in turn is used to generate the OECD series.

Differences in methodologies and terms of references among the sources may arise and explain the given discrepancies for the indicators studied in this Data Review.

201 INCOME DISTRIBUTION DATA REVIEW – SLOVAK REPUBLIC

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD is using data computed from the EU-Survey of Income and Living Conditions (EU-SILC). In the OECD database, income inequality and poverty data are currently available for 2004, 2006, 2007, 2008 and 2009. In the past, the estimates were provided by the Statistical Office of the Slovak Republic (Štatistický úrad Slovenskej Republiky). Since 2011 (5thy data wave), the OECD Secretariat is calculating the indicators in-house while sending the results for verification to the Slovak Statistical Office.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

The number of available reporting on income inequalities and poverty in Slovak Republic is very limited at the national level. We can mention the following:

 The Household and Budget Survey (HBS) is the main survey in Slovak Republic to find data on living conditions and social development. The Statistical Office of the Slovak Republic launched the first Survey in 1957 but data are not available on-line.

 The Microcensus has been used for reporting on household incomes mainly in the 1990s (a microcensus survey took place in 1992 and another one in 1996) though the last microcensus seem to have taken place in 2003 (http://www.statistics.sk/webdata/slov/mikrocen/def_v/zak_v.pdf)

1.2.2 International reporting:

At the international level, we can mention the following reporting:

 EUROSTAT has been computing indicators on inequalities and poverty for Slovak Republic from 2005 (income year: 2004) onwards. Indeed, Slovak Republic entered the European Union in 2004.

 As a consequence, Slovak Republic has been included in the EU-SILC (Statistics on Income and Living Conditions) survey from 2005 onwards (income year: 2004). The EU-SILC is a representative survey of households in Switzerland. This instrument aims at studying poverty, social exclusion and living conditions on the basis of indicators that can be compared at the European level.

 The Luxembourg Income Study Database (LIS) included data for the Slovak Republic for the years 1992 and 1996 only. It is based on the Microcensus survey that is presented in more details in the below table. As data are not available after 1996, the relevance and the interest of this survey for this European country are limited.

 The World Bank data published on PovcalNet, the on-line tool for poverty measurement developed by the Development Research Group of the World Bank (website: http://iresearch.worldbank.org/PovcalNet/index.htm) also includes data for the Slovak Republic. The producer of the data is the Slovak Statistical Office and the survey of reference is the Micro

202 census-Population and Housing Census (except for 1992 – Integrated survey labelled Living Standard Measurement Survey).

The below table presents the main characteristics of those datasets:

Table 30. Characteristics of dataset, Slovak Republic

Microcensus( Luxembourg Income Name Household Budget survey (HBS) EU-SILC from EUROSTAT Study) Responsible Microcensus - Statistical Office of Statistical Office of the Slovak Eurostat the Slovak Republic Republic

Goal Slovak Microcensus To provide information for analyses To study poverty, social To measure the income of the of living standard and social exclusion and living various types of households and to situation of private households, conditions on the basis of get social and demographic especially information on indicators that can be information of the population. The development and structure of their compared at the European main source of funding comes from expenditures and incomes. level. the State budget. Year Since 1958 with an irregular Since 1958 (income year – 1957) Since 2005 (income frequency. year:2004) Provided only for years 1992 and 1996 in the Luxembourg Income Survey (LIS) for Slovak Republic. Data In March (based on incomes from Interviews in households collecting the previous year) using paper questionnaires Data are collected through personal interviews. Covered Wave 1992 - The sampling frame Nationals excluding people from Private Households. population for the survey was the Population remote or inaccessible areas. Census File 1991. The sampling The following types of households are frame includes the total population excluded in the data collection: those of household heads. People living in collective housing (such as long in institutions are also included. term hospitals, prisons, monasteries, Unoccupied units were excluded. military quarters), non-resident The sample frame includes all households of nationals (households geographic areas in the country. of nationals located abroad), Region and district information is diplomatic households in the country, available in the survey. households of other foreigners in the Wave 1996 – Entire Slovak country, armed forces residing in population and territory (1% of all private housing within military base, households). armed forces residing in private housing outside military base. Sample size 17714 unweighted households (no Not provided Minimum sample size is 4250 data was obtained for 1115 households containing 11000 households - wave 1992) individuals (2011 16336 households with 50906 operations). individuals (wave 1996 – Approx. 6068 households (wave 1% of all households in Slovak 2010). Republic). Sampling Two-stage stratified sample Stratified sample One-stage stratified sampling method Some population groups had a was used in EU SILC 2010. higher probability of selection than The proportional number of others; this depends on the size of households was selected by the “counting district” and the size simple random sampling in of the community. individual strata (In the first year of survey, households were selected in to the 4 rotational group on the fact that in each subsequent year of survey one rotational

203 group was excluded and new one was added.) Sampling unit Households Households Households and Individuals Response Non response rate was 6.8 % (wave Not provided Not provided rates 1992) Not provided (wave 1996)

Remark Data are not available on line Annual income of the year prior to the survey Websource http://www.lisdatacenter.org/our- http://portal.statistics.sk/showdoc.do? http://epp.eurostat.ec.europa.e data/lis-database/by- docid=3025 u/portal/page/portal/income_ country/slovak-republic-2/ http://laborsta.ilo.org/applv8/data/SS social_inclusion_living_condi M6/E/462A.html tions/introduction

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The below graph is showing the evolution of Gini coefficients since 1988 as reported by the OECD, the Luxembourg Income Survey, the EU-SILC from Eurostat and the World Bank.

Figure 51. Trends in Gini coefficients at disposal income, Slovak Republic (1988 – 2009)

OECD - Total population Luxembourg Income Survey EU-SILC from Eurostat World Bank 0.40

0.35

0.30

0.25

0.20

0.15

The general trend is that Gini coefficients increased in the 1990s until the mid-2000s (especially in the early 1990s) and slightly decreased between 2005 and 2007. Between 2007 and 2009, trend estimates differ: the Gini increased according to the OECD series and Eurostat, but it decreased according to the World Bank data. That said, all three sources report a similar level in 2009 which is the latest available date. More precisely, Gini coefficients are respectively estimated at 0.246 by the OECD, at 0.259 by the EU-SILC and at 0.26 by the World Bank.

204 The patterns of the share income ratio (S80/S20) trend are broadly similar with the Gini coefficients one. Income distribution equality has increased in the Slovak Republic from 1992 to 2005 as highlighted by the S80/S20 ratios which increased from 2.64 up to 4.39 (World Bank). Afterwards, the trend shows a u-shaped pattern (OECD and Eurostat) or a continuously declining trend. In 2009, latest available data, the 20% of the richest households in Slovak Republic earn 3.8 times more than the 20% of the lowest households (ratio is the same for EU-SILC and OECD). The World Bank is recording a somewhat lower figure with a S80/S20 ratio estimated at 3.58 in 2009. The data of this latter are indeed not marking the recent increase of income inequalities (from 2007).

Figure 52. Trends in S80/S20, Slovak Republic (1988 – 2009)

OECD S80/S20 EU-SILC from Eurostat World Bank 5.0 4.5 4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0

2.1.2 Time series of poverty rates, poverty composition

The below graph shows the poverty rates for the Slovak citizens living with less than 50% of the median equivalised income from 1992 to 2009.

Figure 53. Trends in Poverty rates at 50% median income threshold, Slovak Republic (1992 – 2009)

OECD - After taxes and transfers - Pov Line of 50% Luxembourg Income Study EU-SILC from Eurostat 16% 14% 12% 10% 8% 6% 4% 2% 0%

205 Data on poverty rates can be described and compared only from 2004 onwards. Before this data, poverty rates are available in 1992 and in 1996 and have been only estimated by the Luxembourg Income Study. The World Bank data do not include relative but only absolute poverty estimates.

From 2004 onwards, the OECD data are following the same trend that the EU-SILC ones. The poverty rates with a 50% threshold declined from 8.36% in 2004 (OECD data) down to 6.20% in 2006 (same figures for both the OECD and the EU-SILC). Since 2006 (or 2007 given the fact that EU-SILC is recording a lower poverty rate at 5.60% in 2006), the situation of poverty seems to have worsened. In 2009, 7.7% of Slovaks are living with less than 50% of the median income. Both OECD and EU-SILC point lines are confirming this increase of poverty.

The patterns of the graph regarding poverty rates with a 60% median income threshold are similar to the one here above.

Figure 54. Trends in Poverty rates at 60% median income threshold, Slovak Republic (1992 – 2009)

OECD figures - After taxes and transfers - Pov Line of 60% Luxembourg Income Study EU-SILC from Eurostat 20% 18% 16% 14% 12% 10% 8% 6% 4% 2% 0%

As explained previously, data on poverty rates can be described and compared only from 2004 onwards. Before this data, poverty rates are available in 1992 and in 1996 and have been only estimated by the Luxembourg Income Survey.

From 2004 onwards, the poverty rates with a 60% threshold are following the same trend than the ones with a 50% threshold but the trend is less pronounced. Indeed, the OECD and the EU-SILC point lines seem to have remained pretty stable from 2004 to 2009, respectively from 13.88% to 13.08% for the OECD and from 13.30% to 12.00% for EU-SILC.

From 2008 to 2009, poverty rates, calculated both by the OECD and by the EU-SILC, have worsened. Poverty rates gained approximately one percentage point: between 12% (OECD figure) to 13% (EU-SILC figure) of the total Slovak population is living with less than 60% of the median income in 2009.

2.2. Wages

See Part II of the present Quality Review.

206 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2009 natcur 4137 2016 0 6153 59 671 1964 -916 7931 % av HDI 52.2% 77.6% 0.7% 8.5% 24.8% -11.6% Figure 5 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period.

Figure 5 Trends in shares of public social transfers

4. Metadata of data sources which should explain differences and inconsistencies

There are no major discrepancies to underline. However, we can still mention the following:

 The high increase of poverty rates between 1992 and 1996 as defined by the Luxembourg Income Survey is a bit surprising. With a 50% threshold, the poverty rates has more than tripled (from 2.05% to 7.67%) whereas the poverty rates with a 60% threshold doubled (from 6.25% to 12.86%) between 1992 and 1996. However, no other data are available at these dates to enable comparisons.

 As already mentioned for some other European countries, the S80/S20 ratios as calculated by the Luxembourg Income Survey do not enable comparisons because of a different methodology.

207  The World Bank is not providing the poverty rates below national poverty lines for Slovak Republic conversely to other countries dashboards on the World Bank database.

5. Summary evaluation

Broadly speaking, all the different indicators for Slovak Republic available from OECD, Eurostat and the World Bank are following the same trends between 2004 and 2006-2007. Income inequality and poverty decreased from 2004 to 2006-2007. The last available figures are showing that this trend is reversing between 2007 and 2009 and that Slovak Republic has been facing an increase of income inequalities and of poverty as of today. This latest trend is suggested by OECD and Eurostat data but not by the world Bank data who report a continuous decline.

Beyond the data analysis, the major concerns rely on the too short track record and on the lack of comparative datasets which prevent from undertaking significant data comparisons. The following reasons can be highlighted:

 The availability of data sources for Slovak Republic is very limited. Basically, since 2004 there is only the EU-SILC data available;  The OECD has been compiled statistics for the Slovak Republic since 2004 only and the last available data is 2009. Therefore, the reference period is very short;  The National Office of Statistics is not providing national statistics on social and living conditions in Slovak Republic.

208 INCOME DISTRIBUTION DATA REVIEW – SLOVENIA

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD income distribution and poverty indicators for Slovenia are computed from EU-SILC data from 2004 onwards. The data are computed internally and sent to the Statistical Office of the Republic of Slovenia for verification.

1.2. National reporting and reporting in other international agencies:

 data series based on annual surveys since 2000 (2003 missing).

 LIS database, using annual surveys from the Household Budget Survey for 1997, 1999, and 2004.

 The Statistical Office of the Republic of Slovenia using annual data from EU-SILC since 2004.

The below table presents the main characteristics of those four datasets:

209 Table 31. Characteristics of datasets used for income reporting, Slovenia

OECD reference series National Survey Eurostat LIS database income distribution (income) database Name EU-SILC EU-SILC EU-SILC Household Budget Survey

Name of the Eurostat Statistical Office of the Eurostat Statistical Office of the Republic responsible agency Republic of Slovenia of Slovenia (SORS) and Eurostat Year (survey and 2004 - 2009 2005-2010 EU-SILC 2000-2003 and 1997, 1999 and 2004. income/wage) 2005-2010 surveys representing income for 1999-2002 and 2004-2009. Period over which Annual income for the Annual income for the Annual income for the all Annual income for the all year N- income is assessed all year all year year N-1 1

Covered population All private households All private households All private households and All private households in the who participated in who participated in SILC all persons aged over 16 whole national territory (incl. SILC (except those in collective collective households households and but excl. institutional ones) institutions).

Sample size 9364 households 9364 households 9364 households 3,725 households containing containing 29520 containing 29520 containing 29520 11,303 individuals who individuals (2010). individuals (2010). individuals (2010). completed the interview.

Sample procedure two-stage stratified two-stage stratified two-stage stratified The sample design was stratified sampling design, with sampling design, with sampling design, with two-stage. systematic systematic systematic sampling of sampling of sampling of enumeration enumeration areas enumeration areas areas within each stratum. within each stratum. within each stratum. Response rate About 79% About 79% About 79% The 2004 Household Budget Survey covered 5,268 households, of that 3,725 households responded. Imputation of No missing values, All missing information is fully missing values negative values imputed. treated as suggested in the terms of references Unit for data Individual Individual Individual Household collection

Break in series Missing data for year 2002, representing income for 1999 Web source: http://stats.oecd.org/I http://www.stat.si/doc/ http://epp.eurostat.ec.eur http://www.lisdatacenter.org/wp ndex.aspx?QueryId=2 metod_pojasnila/08- opa.eu/portal/page/portal -content/uploads/our-lis-

6068 025-ME.htm /income_social_inclusion_l documentation-by-si04- iving_conditions/quality/n survey.pdf

ational_quality_reports

210 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among total population has been lower in Slovenia than in any other OECD member and slightly falling from 2004 (0.246) to 2008 (0.236). However, the data from 2009 shows a significant increase compared to 2008 (0.246).

From the four other series of Gini coefficients on disposable income in Slovenia, the EU-SILC series and the Statistical Office of Republic of Slovenia (SORS) series are identical (SORS uses EU-SILC data + methodology), and they show the same hike in 2009. The LIS series shows generally higher levels of inequality than Eurostat series, although data is only available until 2004 and is therefore difficult to compare with the other series.

Figure 55.1 Trends in Gini coefficient (disposable income)

OECD reference series (EU-SILC) Eurostat SORS (EU-SILC) LIS (HBS) 0.30

0.28

0.26

0.24

Gini (at disposable income) (at disposable Gini 0.22

0.20 1995 2000 2005 2010

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference series, the EU-SILC, and SORS, the trends are generally quite similar, although the OECD series shows higher levels. The Eurostat and SORS series are identical since 2004and point to an increase since 2008, following a decrease between 2004 and 2008.

211 Figure 1.2 S80/S20

OECD reference series (EU-SILC) Eurostat SORS (EU-SILC) 4.0

3.5

3.0 S80/S20

2.5

2.0 1998 2003 2008

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the Slovenian population living with less than 50% of the median equivalised income (7327 Euros per year in 2008) has slightly increased from 8.5% in 2004 to 8.7% in 2009.

The series from Eurostat shows trends that are similar to the OECD reference series in both poverty rates and Child poverty rates, but it also shows lower levels throughout. LIS series differ markedly from Eurostat series.

Figure 2.1 Trends in poverty rates

OECD reference series (EU-SILC) Eurostat LIS (HBS)

12%

10%

2% Poverty rate (at 50% median income) median (at 50% rate Poverty

0% 1994 1999 2004 2009

212 Figure 2.2 Trends in Child poverty rates

OECD reference series (EU-SILC) Eurostat LIS (HBS)

12%

10%

2% Child Poverty rate (at 50% median income) median 50% (at rate Poverty Child 0% 1995 2000 2005 2010

2.2 Wages

See Part II of the present Quality Review

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 32. Shares of income components in total disposable income, OECD reference series

Self Survey Year Unit Wages Capital Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2008 natcur 13,660 352 1,187 3,852 -4,452 14,653 % av HDI 93% 2% 8% 26% -30%

213 Figure 3. Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and the other EU-SILC series :

The OECD reference series (as well as the LIS series) use the square root of household size, whereas the EU-SILC series and SORS series use the OECD modified equivalence scale (1.0 to the first adult, 0.5 t o the second and each subsequent person aged 14 and over, 0.3 to each child aged under 14).

5. Summary evaluation

The trends of the OECD reference series and of EU-SILC are generally quite consistent, although the OECD reference series exhibits higher levels overall. This may be due to the difference equivalence scales used (square foot of household size for the OECD series; OECD modified equivalence scale for EU-SILC) since the samples used are identical.

214 INCOME DISTRIBUTION DATA REVIEW – SPAIN

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD series for Spain starts back in the 1980’s and is based on three different sources:

 The Continuous Survey of Household Budget (“Encuesta Continua de Presupuestos Familiares”) for the years 1985, 1990 and 1995

 The European Community Household Panel (ECHP) for the years 1995 and 2000 (data remains comparable for these two years despite some slight changes in the methodology of the ECHP)

 The European Union Survey on Income and Living Conditions (EU-SILC) for the years 2004, 2006, 2007, 2008 and 2009.

Due to the change in source in 1995, data for 1985 and 1990 have been interpolated (“spliced”, with the overlapping year 1995) and the OECD Datacube provides series for both the original and the adjusted values. However, there is a strict break in series due to the change in source to EU-SILC in 2004, with no overlapping year with two data sources available. Data prior to 2004 are strictly not comparable with data from the 2004 onwards. This break between 2000 and 2004 is shown in the Figures below. The figures also show the original (“unspliced”) series.

1.2. National reporting and reporting in other international agencies:

Alternative sources are available for the computation of income distribution and poverty figures, both at the national and European level. At the national level, the following surveys are undertaken by the National Statistics Office (Instituto Nacional de la Estadistica):

 The Living Conditions Survey (LCS) (Encuesta de Condiciones de Vida (ECV)) It presents the same data than EU-SILC as it has been set up as the national survey underlying the EU-SILC User Database (UDB). This survey has been launched in 2005 (income year: 2004) and is the one used currently for official figures in Spain. Produced annually and household based, it has been preceded by the PHOGUE (see below) carried out during the 1994-2001 period. Both surveys have similar characteristics and objectives.

 The European Union Households Panel (ECHP) (Panel de hogares de la Unión Europea (PHOGUE)). This survey has been carrying out during the period 1994-2001. It has facilitated the European Commission having a sound statistical instrument available to follow up social cohesion in the relevant territory, the study of population needs and the impact of social and economic policies on households and persons, as well as for the design of new policies. Once the eight cycles have been completed, a global filtering of information and revision of the tables corresponding to the years 1996-1999 has been carried out. This survey is now replaced by the Living Conditions Survey (LCS) (=Encuesta de Condiciones de Vida (ECV)). In 2000 there were two ECHP samples, one with 5.132 and another with 15.614 households (c.f., http://webs.uvigo.es/microsimulacion/papers_pdf/Fuenmayor_Granell_PHOGUE- ECV_paper.pdf).

215  The Households Budget Survey (HBS) (Encuesta de Presupuestos Familiares (EPF)) This survey has been launched in January 2006 to replace the previous survey called “Encuesta Continua de Presupuestos Familiares” (ECPF). This survey provides annual information on the nature and destination of consumption, as well as on a range of features relating to household living conditions. The survey provides estimates of annual consumption for the entire country and the Autonomous Communities, and of consumption in physical amounts of certain food items for the country as a whole.

 The Household Budget Continuous Survey (Encuesta Continua de Presupuestos Familiares (ECPF)) This survey, started by INE in January 1985, used to provide quarterly and annual information on the origin and amount of households incomes, and the way they are used in several consumption expenditures. This survey is available from 1985 to 2005.

At the European level, the following studies are available;

 The Luxembourg Income Study (LIS) Data on LIS are available for the years 1980, 1990, 1995, 2000 and 2004. The LIS is based on the Family Expenditure Survey for 1980 and 1990, on the PHOGUE for 1995 and 2000 and then on EU-SILC for 2004. As a result, OECD series are in based on the same sources as for the LIS for the years of availability.

The table below summarises the main characteristics of these several sources:

Table 33. Characteristics of datasets, Spain

Name Household Household Budget Budget Survey European Union The Living Conditions Continuous (HBS) Households Panel Survey (LCS) Survey – EU-SILC from (=Encuesta de (EUHP) (=”Panel de (=Encuesta de (Encuesta EUROSTAT Presupuestos hogares de la Unión Condiciones de Vida - Continua de Familiares Europea” (PHOGUE) ECV) Presupuestos (EPF)) Familiares) Name of the National National National Statistical National Statistical Eurostat / responsible Statistical Office - Statistical Office Office - Instituto Office - Instituto National agency Instituto Nacional - Instituto Nacional de la Nacional de la Statistical Office de la Estadistica Nacional de la Estadistica (INE) Estadistica (INE) (INE) Estadistica (INE) Goal To provide To provide Producing Systematic production To study information on information on comparable of Community statistics poverty, social spending patterns consumption information across on income and living exclusion and for the Spanish structure and the Member States conditions, that include living conditions retail price index living conditions on income, work and transversal and on the basis of employment, poverty longitudinal indicators that and social exclusion, data that are can be housing, health and comparable and compared at many other diverse updated on income, the the European social level and composition level. indicators concerning of poverty and social living conditions of exclusion, on a national private households and European level. and persons Year 1985-2005 2006-continuing 1994-2001 Since 2005 Since 2005 (income year 2004) Data Annually Annually Annually Annually Annually.

216 collecting Data are collected from April to June Covered All households All households Private households in All private households All private (non- population except those except those the whole national in Spain and overseas group , non- living in private living in private territory (with the territory institutional) institutions institutions exception of Ceuta households in and Melilla) Spain and overseas territories (Ceuta, Melilla, Canarias). Sample size The sample size is 24000 7,206 households, 13,026 interviewed The theoretical approximately households per 4,966 completed households sample includes 24,000 year households (for 1994) about 16,000 households per dwellings year. Each distributed in household 2,000 census remains in the sections. The sample for two effective consecutive sample years, with half of includes 13,026 the sample interviewed renewed each households year. (wave 2005, survey year). Sampling Two stage Two stage Two-stage sampling: The sample is rotating, Stratified survey method stratified stratified first a with a quarter of the First stage unit: sampling design sampling design sample area was sample renewed each area framework selected, and cycle (year). Each formed by the then, within this area, subsample is selected relation of a using a twostage existing census building design with first stage sections used in object/housing unit unit stratification is the 2003 was selected used. The first stage is Municipal formed by census Register of sections (grouped into Inhabitants. 6 strata Second stage according to size of unit: list of main municipalities within family dwellings each of the 18 in each of the Autonomous sections Community) selected selected within each stratum with a probability proportional to their size, and the second stage by main family dwellings selected with equal probability (via random start systematic sampling). Sampling unit Households and Households and Households and Households and Households and individuals individuals individuals individuals Individuals

217 Response 63% n.a. n.a. 82% n.a. rates

Imputation Hot deck and n.a. None n.a. deterministic

Remark Annual income of the year prior to the survey Websource http://www.ine.e http://www.ine. http://www.lisdatace http://www.lisdatacente http://www.lisda s es nter.org/ r.org/ tacenter.org/wp- content/uploads/ our-lis- documentation- by-es04- survey.pdf

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

Gini series are available from the Luxembourg Income Study, EU-SILC, and for the ECV.

Figure 56. Gini coefficients, Spain (1980-2010)

OECD Luxembourg Income Survey Encuesta de condiciones de vida (ECV) Eurostat EU-SILC from Eurostat 0.40

0.35

0.30

0.25

218 From 2004 to 2010, the OECD series tracks very closely the EU-SILC, unsurprisingly as the OECD one is based on EU-SILC. Figures from the ECV track also those two series, as this survey is the national basis for Eu-SILC. The LIS series displayed an almost similar trend than the OECD one, while the sources are not fully comparable.

Unsurprisingly, the S80/S20 ratio series display the same features, while one additional series from the EUHP-PHOGUE is included:

Figure 57. S80/S20 ratio, Spain (1980-2010)

OECD S80/S20 Eurostat Eurostat (EU-SILC) Encuesta de condiciones de vida (ECV) Panel de hogares de la Unión Europea (EUHP) 8

2.1.2 Time series of poverty rates, poverty composition

Time series on poverty rates are available from almost all the sources presented above. They are based on two standard poverty threshold: 50% and 60% of the median equivalised disposable income.

Figure 58. Poverty rates with a poverty line of 60%, Spain (1980-2011)

OECD Luxembourg Income Survey Panel de hogares de la Unión Europea (EUHP) Encuesta Continua de Presupuestos Familiares (ECPF) Encuesta de condiciones de vida (ECV) EU-SILC from Eurostat

24%

22%

20%

18%

16%

14%

12%

10%

219 Figure 59. Poverty rates with a poverty line of 50%, Spain (1980-2010)

OECD Luxembourg Income Survey Eurostat EU-SILC from Eurostat 18%

16%

14%

12%

10%

6% 1980 1982 1984 1986 1988 1990 1992 1994 1995 1997 1999 2001 2003 2005 2007 2009 2011

Figure 60. Chile Poverty rates with a poverty line of 50%, Spain (1980-2010)

OECD - Age group 0-17 - Pov Line 50% INE - Panel de hogares de la Unión Europea INE - Encuesta Continua de Presupuestos Familiares. Base 1997 INE - Encuesta de condiciones de vida EU-SILC from Eurostat _ Children - Pov line 50% Luxembourg Income Study 30%

25%

20%

15%

10%

For the 60% and 50% threshold, poverty rates display very similar trend and values, LIS series included. Regarding child poverty, discrepancies are more important between national and OECD sources despite the fact that the definition of the concept is supposed to be similar. Differences in equivalence scales and data treatment (e.g. with regard to negative incomes) may play a role.

2.2 Wages

See Part II of the present Quality Review

220 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2009 natcur 8723 4442 20 13185 567 1366 4144 -2725 16537 % av HDI 52.7% 79.7% 3.4% 8.3% 25.1% -16.5%

Figure 6 Trends in shares of public social transfers

3. Metadata of data sources which could explain differences and inconsistencies

Regarding income inequality indicators, the OECD time-series prior to 2004 can be compared with the LIS data series as both use a similar methodology and the same data source. For the years after 2004, OECD series broadly match with estimates published by Eurostat, based on EU-SILC. Regarding the comparison on poverty rates, figures for the total population compare well. Only child poverty rates are somewhat different and further investigation into methodological differences is needed to account for such differences.

221 5. Summary evaluation

All in all, there is a high degree of similarity between national, LIS, EU-SILC and OECD series. This comes from the integrated way in which the series are produced: the national sources are used to generate EU-SILC, which in turn is used to generate the OECD series. At each stage, few methodological changes occur, which leads to the similarities observed.

222 INCOME DISTRIBUTION DATA REVIEW – SWEDEN

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD Income Distribution Data for Sweden are computed by Statistics Sweden and based on the Income Distribution Survey (IDS). Data from IDS is available since 1975. Due to a change in the household definition in 1995 (move from fiscal to demographic household definition), data prior to 1995 have been interpolated for the current definition.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

Income distribution and poverty indicators for Sweden are also available from Statistics Sweden which is also based on the IDS, which is a sample survey carried out every year. The population for the survey consists of all households and people who were registered in the population register in Sweden at some point during the survey year (income year). The Total Population Register is used for the sampling frame. The sample is made up of people aged 18 or over. Up to and including the income year 2000, the sample was coordinated with Statistics Sweden's Longitudinal Individuals Database (LINDA). The coordination with LINDA provides the possibility to follow sample persons and their households over several years.

1.2.2 International reporting:

The Luxembourg Income Study (LIS) included Sweden in years 1987, 1992, 1995, 2000 and 2005. It is also based on IDS.

EUROSTAT has been computing indicators on inequalities and poverty for Sweden from 2001 (income year 2002) onwards using EU-SILC.

Table 1 presents the main characteristics of the different sources

223 Table 34. Characteristics of datasets used for income reporting, Sweden

OECD reference series LIS database Statistics Sweden Eurostat income distribution database Name Income Distribution Survey Income Distribution Income Distribution EU-SILC (IDS) Survey (IDS) Survey (IDS) Name of the Statistics Sweden Statistics Sweden Eurostat responsible agency Statistics Sweden (www.scb.se) (www.scb.se) (www.scb.se)

Year (survey and 1975, 1983, 1991, 1995, 1987, 1992, 1995, 2000, 1975-2010 2002-2011 survey years income/wage) 2000, 2004, 2008. Due to a 2005 representing income year for change in the household 2001-2010 (2002 income year definition in 1995, data missing) prior to mid 90s have been interpolated for the current definition. Period over which All income data refers to Annual income The population for the Annual income N-1 income is assessed the income earned during survey consists of all the current year. Due to the households and people taxation process data is who were registered in available with about 18 the population register months delay. in Sweden at some point during the survey year (income year). Covered population Individual; most individual variables are then household aggregated to the household level Sample size 16268 households 2010: 7173 households 2008: 16197 households containing 36918 individuals (2005) Sample procedure Cross-sectional household Cross-sectional longitudinal one stage systematic sampling survey with integrated design register data. Response rate In 2008 the non-response rate was 34,5 percent. Imputation of No missing incomes, missing values negative incomes are included. Unit for data Resident population in Household individual Household collection private households Break in series No No No No

Web source: http://www.oecd.org/els/s http://www.lisdatacente http://www.scb.se/Page http://epp.eurostat.ec.europa. ocialpoliciesanddata/incom r.org/our-data/lis- s/TableAndChart____16 eu/portal/page/portal/income edistributionandpovertydat database/by-country/ 3547.aspx _social_inclusion_living_conditi afiguresmethodsandconcep ons/quality/national_quality_r ts.htm eports

224 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The OECD reference series for Gini coefficients of disposable income are quasi identical to the Statistics Sweden series based on IDS. The LIS series, which is also based on IDS, exhibits similar trends although the levels are slightly higher. The EU-SILC series, although similar until 2005, then shows contrasting trends with fluctuating levels, as opposed to a steady increase for the OECD and Statistics Sweden series. The same observation applies to an alternative inequality indicator, the S80/S20 share.

Figure 61.1. Trends in Gini coefficient (disposable income)

OECD reference series (IDS) LIS (IDS) Statistics Sweden (IDS) Eurostat (EU-SILC) 0.30

0.25

0.20 Gini (at disposable income) (at disposable Gini

0.15 1990 1995 2000 2005 2010

Figure 1.2. S80/S20

225 2.1.2 Time series of poverty rates

The OECD reference series shows increasing levels of poverty rates and child poverty rates since 2004. The LIS series, despite the fact that it is also based on IDS shows higher levels of poverty rates throughout and even shows a contrasting trend for the total poverty rate between 1995 and 2005, with a slight decline, while the OECD reference series shows a steady and constant increase. The EU-SILC series is more fluctuating and, while the general increase it exhibits is concomitant to the OECD reference series, the year 2010 shows contrasting trends with the EU-SILC showing declining levels of both poverty and child poverty rates while the OECD reference series shows increasing levels. Furthermore, for the EU- SILC series, there is an unexplainable spike in 2006.

Figure 2.1. Trends in poverty rates

Figure 2.2. Trends in child poverty rates

226 2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series, and for the EU-SILC data. The OECD data have a higher coverage of capital income, at the (proportionate) expense of the shares of all other income sources.

Table 35. Shares of income components in total disposable income, OECD reference series

Figure 3. Trends in shares of public social transfers

227 4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Methodological differences between the OECD reference series and the other income series:

5. Summary evaluation

OECD series of income inequality and poverty match well with series published by Statistics Sweden and rather well with the LIS series. The OECD series and EU-SILC series are diverging in the last couple years in both inequality and poverty rates. The remaining and the earlier discrepancy still needs to be explained.

228 INCOME DISTRIBUTION DATA REVIEW – SWITZERLAND

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD is using the two types of sources:

 From 2002 to 2005, the OECD were provided data from the Income and Consumption Survey which was computed by the Federal Statistical Office. Please, note that income year surveys are labelled as 2000-2001 and 2004-2005 because income data refer to averages of two consecutive years (data have been merged in order to increase significance of results). For the purpose of this study and to enable more comparisons, please note that the OECD data for years 2000-2001 and 2004-2005 have been respectively reported as 2000 and 2004.

 From 2008 onwards, the OECD has been using the EU-Survey of Income and Living Conditions (EU-SILC) under the supervision of Eurostat.

In the OECD database, income inequality and poverty rates are currently available only for 2000- 2001, 2004-2005 and 2008. The change of survey in 2008 (from the Income Consumption Survey to EU- SILC one) is considered as a strict break. It means that the OECD data cannot be strictly compared before and since 2008.

1.2. National reporting and reporting in other international agencies:

1.2.1 National reporting:

In addition to the OECD time-series, the following reportings are available on income inequalities and poverty for Switzerland at a national level:

 The Income and Consumption Survey (= Enquête sur les Revenus et la Consommation – ERC) is the main survey in Switzerland to find data on income distribution and poverty. The Swiss Federal Statistical Office launched the first Survey on Income and Consumption in 1990 and the second one took place in 1998. Since 2000, this Survey has been taking place on an annually basis. This survey changed name in 2008 and became “Household Budget Survey” (="Enquête sur le budget des ménages" (EBM/HBS)).

Please, note that there was a change in the weighting model in 2003 which led to a recalculation of the indicators estimated in the ERC for the years 2000, 2002 and 2003 which enable to compensate for a possible bias in the selection of households. The HBS results are therefore based on a household structure which accurately represents the permanent resident population in Switzerland.

The below template present the main differences between the previous and the new weighting model in 200324:

24 Anne Cornali Schweingruber, Ruedi Epple, Ueli Oetliker, Sylvie Rochat, « Une nouvelle méthode pour l’Enquête sur les Revenus et la Consommation (ERC) », dans la série « Statistique de la Suisse », Office Fédéral de la Statistique, Neuchatel 2007.

229

In 2006, there was also a contents and process optimisation25.

The main differences between the EU-SILC survey and EBM (formerly ERC) can be resumed as follows:

 Description : Statistique sur les revenus et conditions de vie (SILC) ; Enquête sur le budget des ménages (EBM anciennement ERC)

 Type de questionnaire : CATI ménage et individuel ; Questionnaire papier accompagné de CATI. Ménage soutenu par un enquêteur pour remplir le questionnaire.

 Nature de l’enquête : Panel de ménages rotatif sur 4 ans ; Echantillon mensuel de ménages

 Période de référence du revenu Annuel (t-1 pour enquête SILC t) ;Mensuel pour les revenus réguliers et annuel pour les revenus ponctuels (primes, 13ème mois, …)

 Période d’enquête : De février à juillet ; Au cours de toute l’année

 Nombre de ménages participant par année : 6500 ; 3300

 Taux de participation : Plus de 65% ; Moins que 35%

 Utilisation de registres administratifs : Pour la consolidation des revenus et la réduction de non- réponse ; Non

1.2.2 International reporting:

In addition to the OECD time-series, the following reportings are available on income inequalities and poverty for Switzerland at an international level:

 EUROSTAT has been computing indicators on inequalities and poverty for Switzerland from 2008 (income year: 2007) onwards. Switzerland signed the bilateral agreement on statistics with the EU in 2007 and started taking part in the annual statistics program of the European Union together with all members of the Union. This explains why Eurostat data are not available before

25 With the revision of the Household Budget Survey in 2006, the definitions of gross household income and disposable income were adapted to meet new international standards. For example, sporadic income is no longer included in gross income (nor in disposable income).

230 this year for Switzerland. In 2010 Switzerland became a full member of the European Statistical system (ESS).

 As a consequence, Switzerland had been included in the EU-SILC (Statistics on Income and Living Conditions) survey from 2008 onwards (income year: 2007). The EU-SILC is a representative survey of households in Switzerland. This instrument aims at studying poverty, social exclusion and living conditions on the basis of indicators that can be compared at the European level. Every year, both cross-sectional data (pertaining to a given time or a certain time period) and longitudinal data (pertaining to individual-level changes over time, observed periodically over a four year period) are collected.

 The Luxembourg Income Study Database (LIS) included Switzerland in 1982 and relied on statistics produced by the Swiss Federal Statistical Office. Data are available for years 1982, 1992, 2000, 2002 and 2004. For 1982, the LIS relied on the Swiss Income and Wealth Survey. In 1992, it used the Swiss Poverty Survey. Afterwards, the LIS referred to the Income and Consumption Survey (EVE/ERC) which was used for years 2000, 2002 and 2004.

Table 1 presents the main characteristics of those four datasets:

Table 36. Characteristics of datasets used for income reporting, Switzerland

Luxembourg Income Study Income and Consumption OECD and Eurostat Name Swiss Income and National Poverty Survey (Enquête sur les (EU-SILC) Wealth Survey Survey Revenus et la Consommation - ERC) Name of the Institute of Institute of Economics Federal Statistical Office Eurostat / Federal responsible Economics (=Volkswirtschaftliches Statistical Office agency (=Volkswirtschaftlich Institut), University of es Institut), University Bern of Bern Goal To measure income To ‘picture’ the extent To provide information on To study poverty, social distribution in of poverty and to collect patterns of consumption exclusion and living Switzerland. information about the and income of households conditions on the basis of living as well as to determine the indicators that can be conditions and income. yearly rate of price compared at the evolution. European level. It changed name in 2008 to become the Household Budget Survey (HBS). Year 1982 and only this 1992 1990, 1998 and every year Since 2008 year since 2000 Data n/a n/a Annually since 2000. Annually. From February collecting From January to December. to July

Covered The main sampling The private households Permanent resident population frame consists of a list residing permanently within population in private of registered voters the borders of Switzerland. households. (i.e. Swiss citizens). Border residents, foreign Social exclusion and An additional sample tourists, and collective housing condition of foreign nationals households (e.g. prisons) information is collected at holding permanent are not taken into account in household level while residence was this survey. labour, education and randomly selected health information is from a central register obtained for persons aged of foreign nationals. 16 and over. Foreign household

231 heads without permanent residence permits and mentally disabled persons were not included in the sampling frame. Military personnel and people living in institutions were included in the sampling frame. Sample size In 1982, 7036 tax In 1996, 6301 persons. In 2004, 3,270 households The minimum effective units for which data containing 7,993 sample size is 4250 was collected. 6055 of individuals who completed households. In 2010, the these were comprised the interview. actual sample size was of Swiss citizens and Since 2008, when it became 10547 households. 981 by foreigners. HSB, about 3000 households are taking part each year. Sampling Stratified The sample design used The survey was conducted Stratified random method for the survey was a on the basis of 12 random sampling design stratified sample, in monthly samples, stratified which persons with a according to the seven low income and persons grand regions of above 60 years of age Switzerland. Sample were over represented. households are chosen at The sample is self- random from the register of weighting. private telephone numbers. The HBS is conducted by means of telephone interviews and written questionnaires. Note: To obtain a sufficient number of households in each region, an oversampling of the Tessin region has been made. Sampling unit Tax Units Households and Households and Individuals Households Individuals Response For 128 (1.8%) of the Response rates The response rate over 12 The non-response rate rates 7,036 cases income correspond to a months collection period was 25% approx. for the data is not available. percentage of 36.7%. was on average around total sample (wave 2010) Response rate is 33%. around 98%

Remark Remark on the weighting Annual income of the change model is explained year prior to the survey in section 1.2 Websource http://www.lisdatacen http://www.lisdatacente http://www.lisdatacenter.or http://www.bfs.admin.ch/ ter.org/wp- r.org/wp- g/wp-content/uploads/our- bfs/portal/en/index/infoth content/uploads/our- content/uploads/our-lis- lis-documentation-by-ch04- ek/erhebungen__quellen/ lis-documentation-by- documentation-by- survey.pdf blank/blank/silc/01.html ch82-survey.pdf ch92-survey.pdf http://www.bfs.admin.ch/bf s/portal/en/index.html

232 2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

The Federal Office of Statistics estimated Gini coefficients for Switzerland from 1998 in the study entitled “Survey on Income and Consumption (“Enquête sur les Revenus et la Consommation – ERC). This study calculated three types of Gini:

 The Gini coeffient based on Gross Income

 The Gini coefficient based on Disposal Income

 The Gini coefficient based on short term disposal income

In this Data Review and for the purposes of comparisons, only the Gini coefficient based on disposal income has been included. The below graph is showing the evolution of Gini coefficients since 1982 as reported by the OECD, the Luxembourg Income Survey, the national Statistical Office and Eurostat.

Figure 62. Trends in Gini coefficients, at disposable income

OECD - Reference series CSO - Income and Consumption Survey/Household Budget Survey Luxembourg Income Study Eurostat (EU-SILC) 0.40

0.35

0.30

0.25

0.20

0.15

1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 2006 2007 2008 2009 2010 1982 1983 1984 1997 1998 1999 2000 2001 2002 2003 2004 2005

Over the period (1982 – 2009), income inequalities remained broadly stable with a Gini coefficient contained between 0.26 and 0.3, a coefficient which was slightly below the OECD average. More precisely, from 1982 up 2004, according to LIS data, the Gini coefficient slightly declined. A Switzerland Statistics Office’s publication26 looking at the income distribution from 1998 to 2004 highlighted the fact that there was a slight decrease of inequalities over this time period and that the distribution between gross income and disposal income were pretty similar.

26 Caterina Modetta, Bettina Müller, “ Inégalité des revenus et redistribution par l’Etat - Composition, répartition et redistribution des revenus des ménages privés ”, dans la série Statistique de la Suisse, Office Fédéral de la Statistique, Neuchatel 2012.

233 OECD data between 2000 and 2004 are totally matching with these other surveys available for this period, ie. the Luxembourg Income Survey and the Income and Consumption Survey. The spread between the different available data is below a 0.01 point of difference with Gini coefficients estimated around 2.7- 2.8. After 2004, Gini coefficients slightly increased to reach level of 0.3 in 2009. OECD data consider a strict break between 2004 and 2008. The OECD is recording a level of 0.302 in 2008 which is the same as the Eurostat figure. The National Statistics Office is suggesting an increase over this period between 2004 and 2009 (mostly due to a surge in the year 2006).

This relative steadiness of Gini coefficients is confirmed by the trend regarding the share of income ratios which are calculated below both for S80/S20 and for d90/d10.

The OECD S80/S20 ratios in 2000 and 2004 match with the ones provided by the Income and Consumption Survey while being slightly higher. These ratios remained pretty stable over the period and in 2004, the average income of those in the highest income quintile was 4.2 times that of those in the lowest quintile. In 2008, this ratio increased up to 4.6 in line with the figures from Eurostat. Indeed, the ratios estimated by Eurostat have been always a little higher than the ones calculated by the National Statistical Office.

Figure 63. Trends in S80/S20 ratios

OECD S80/S20 CSO - Income and Consumption Survey/Household Budget Survey Eurostat 6

1982 1983 1985 1987 1989 1990 1992 1994 1996 1997 1999 2001 2003 2004 2006 2008 2010 1984 1986 1988 1991 1993 1995 1998 2000 2002 2005 2007 2009

The evolution of the median income can be compared between the OECD time-series and the EU- SILC/Eurostat and the Living Income Survey. The data are completely matching and this can be easily explained. Indeed, for years 2000 and 2004, both the OECD and the Luxembourg Income Survey used the Income and Consumption Survey from the National Statistics Office to estimate the median equivalised income. In 2008, the OECD used Eurostat/EU-SILC data so it makes sense that EU-SILC and OECD median equivalised income are alike.

234 Figure 64. Trends in median equivalised income

2.1.2 Time series of poverty rates

From 1982 to 2009, poverty rates remained pretty stable in Switzerland in line with the findings mentioned in the previous income distribution part. OECD time-series between 2000 and 2004 can be compared with the Luxembourg Income Survey whereas OECD data in 2008 can only be compared with Eurostat.

According to the OECD, the poverty rates with a 50% threshold slightly increased from 2000 (7.5%) to 2004 (8.7%) while remaining always below the OECD average (11% in the mid-2000s27). The OECD figures are broadly matching with the Luxembourg Income Study data for these years.

In 2008, the poverty rate with a 50% threshold was recorded to a higher level at 9.26% but we cannot conclude to an increase of poverty in Switzerland given the fact that there is a change of survey in 2008 for the OECD dataset and that figures are not provided by the National Statistical Office (Households Budget Survey do not provide this data).

The below graph shows the change in poverty rates for the Swiss population living with less than 50% of the median equivalised income over the given period.

27 “Growing Unequal? Income Distribution and Poverty in OECD Countries (2008)”, OECD, Paris.

235 Figure 65. Trends in poverty rates at 50% median income threshold

The patterns of the graph regarding poverty rates with a 60% median income threshold are similar to the one hereabove. According to the Luxembourg Income Study, poverty rates remained pretty stable from 1982 to 2000 and increased between 2000 and 2004. Both the OECD and the Luxembourg Income Survey are confirming this trend as the share of the Swiss population living with less than 60% of the median equivalised income rose from 13.34% to 15.22% (OECD figures). Even though the Luxembourg Income Study data is recording a smaller increase than the OECD one, the trend is similar between the both datasets. Despite this increase, the poverty rates in Switzerland remained below the average of the OECD member states which was estimated at 17% in the mid-2000s28.

The poverty rates provided by Eurostat for the period 2007-2009 are slightly higher than those estimated by national CSO on the same data. However, it is hazardous to do any comparison since the Luxembourg Income Survey and EU-SILC/Eurostat did not calculate poverty rates for any years in common. Furthermore, as mentioned previously, OECD data between 2004 and 2008 cannot be compared neither. In 2008, according to the OECD, the share of the population living with less than 60% of the median income was estimated at 16.07%.

In 2009 (year of survey: 2010), 14.2% of the Swiss population lives in relative poverty with a 60% threshold according to Eurostat data. Some types of households are more vulnerable than others, especially individuals living alone with an equivalised disposal income below 2400CHF per month or households with two children with an equivalised disposal income below 5000 CHF per month29.

In 2009 and according to EU-SILC, it is interesting to note that the share of Swiss population living with 50% of the equivalised median income is estimated at 7.6% whereas this rate is nearly doubled to 14.2% when the threshold is fixed at 60%. This means that a significant number of people are between the 50% and the 60% threshold and could be out of the at-risk-of poverty if they could get slightly higher earnings30.

28 “Growing Unequal? Income Distribution and Poverty in OECD Countries (2008)”, OECD, Paris. 29 Swiss Statistics Office, Press Release n° 0351-1112-80, « Les ménages avec enfants ont plus de difficultés à faire face à une dépense imprévue », 15-12-2011, in French only, 30 Swiss Statistics Office, Press Release n° 0351-1112-80, « Les ménages avec enfants ont plus de difficultés à faire face à une dépense imprévue », 15-12-2011, in French only,

236 Figure 66. Trends in poverty rates at 60% median income threshold

Child poverty rates are a little higher than those for the total population. The below graph is presenting children poverty rates from 1982 to 2009. However, the trend is similar than for the poverty rates for the total population.

For the period 2000-2004, the OECD children poverty rates are similar with the ones estimated by the Luxembourg Income Survey. In 2004, the children poverty rate was estimated at 9.43% according to the OECD, one percentage point higher than for the total population. In 2008, this rate rose to 9.26% but, as explained previously, comparisons are not possible between 2004 and 2008. This OECD rate is very similar than the rates calculated by Eurostat and EU-SILC.

Figure 67. Trends in Child poverty rates (50% median income threshold)

2.2 Wages

See Part II of the present Quality Review.

237 3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2. Shares of income components in total disposable income, OECD reference series

Average income Average income K SE TR TA HDI

Survey Year Unit EH ES EO Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery 2008 natcur 43265 11052 4442 58760 3793 6676 12662 -21833 60058 % av HDI 72.0% 97.8% 6.3% 11.1% 21.1% -36.4%

Table 3 below lists the detailed income components which have been included in the OECD series for Switzerland 2000 and 2004.

238 Table 3. Definition and classification of income components from ERC/EBM

EH Revenus issus d'activités salariées de la personne de référence ES Revenus issus d'activités salariées du partenaire de la personne de référence EO Revenus issus d'activités salariées des autres personnes du ménages 1 Salaire (mois 1 à 12) des employés, sans suppléments, avant déductions (brut) 2 Salaire (13 et 14ème) des employés, sans suppléments, avant déductions sociales (brut) 3 Participations aux bénéfices, tantièmes (bruts) 4 Indemnités de résidence (brut) 5 Gratifications, primes de fidélité et au rendement (brut) 7 Pourboires encaissés (nets) 8 Indemnités pour travail irrégulier ou pénible et jetons de présence à tout employé (brut) 9 Revenus des membres des autorités de la Confédération, des cantons et des communes (brut) 10 Indemnités de départ, de vacances ou pour jours fériés (brut) 11 Prestations en nature de l'employeur K Revenus de la fortune 22 Recettes (nettes) issues de la location de terrains 23 Intérêts, dividendes 24 Rentes issues d'assurances-vie (fortune) 42 Rentes supplémentaires privées, issues de transferts 50 Indemnités des assurances privées non obligatoires SE Revenus issus d'activités indépendantes et revenus de la location 6 Salaire des employés de leur propre entreprise, sans suppléments, avant déductions (brut ) 12 Prélèvements dans la caisse de l'entreprise propre, utilisés pour les besoins du ménage (net) 13 Paiements directs d'acquisitions du ménage par la caisse/les comptes de l¿entreprise propre 15 Revenus (nets) issus d'activités économiques informelles du ménage 16 Autoproduction du ménage pour sa propre consommation (jardin, clapier, entreprise propre, etc.) 17 Sous-location (brute) du logement principal loué, à des tiers 18 Sous-location (brute) des résidences secondaires louées, à des tiers 19 Location (brute) du logement principal en propriété, à des tiers 20 Location (brute) des résidences secondaires en propriété, à des tiers 21 Location (nette) de biens immobiliers en possession du ménage, uniquement destinés à l'usage par des tiers TR Prestations sociales odsb 25 Rentes ordinaires de l'AVS/AI odsb 26 Prestations complémentaires de l'AVS/AI odsb 27 Rentes extraordinaires de l'AVS/AI odsb 28 Allocations pour impotents de l'AVS/AI odsb 29 Rentes de caisses de pension (PP) UB 30 Indemnités de l'assurance-chômage (IAC) OIDB 31 Indemnités journalières des assurances accidents et maladie professionnelles (PM AAMP) OIDB 32 Indemnités journalières des caisses maladie et accidents privées (PM AM) FCB 33 Allocations familiales fédérales, pour les agriculteurs (AFf) FCB 34 Allocations familiales liées au besoin (Afc)(allocations de maternité, de naissance, pour enfant) UB 36 Indemnités pour chômeurs en fin de droit OCB 37 Subsides pour le paiement des primes de l'assurance-maladie (SPAM) HB 38 Subsides pour le paiement du loyer (SPL) OTH 39 Indemnités pour perte de gain durant le service militaire et de protection civile (APG) OTH 40 Prestations monétaires de l'assurance militaire (PM AMil) OTH 41 Autres prestations cantonales et communales: assistance sociale (ASS), aide aux victimes d'infractions, prestations supp OTH 47 Bourses d'études, subsides à la formation OTH 49 Pensions alimentaires TA 821 Assurances-vieillesse et survivants (AVS)/-invalidité (AI)/-perte de gain (APG) 822 Assurance-chômage (AC) 823 Assurance-accidents professionnelle (LAA) 824 Caisse de pension (LPP) 825 Autres assurances sociales des personnes actives 829 Assurance-maladie de base 830 Assurance hospitalière complémentaire 831 Autres assurances-maladie et -accidents complémentaires 843 Impôt fédéral direct 844 Impôts cantonaux sur le revenu et la fortune 845 Impôts communaux, paroissiaux et autres sur le revenu et la fortune 846 Impôts cantonaux, communaux, paroissiaux et autres confondus 847 Impôts à la source Figure 8 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show somewhat different trend, with the OECD SOCX data suggesting more stability.

239 Figure 8 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

Only minor differences appear between the OECD reference series and the results from different other data sources for Switzerland. Some data between the Eurostat database and the nationally published EU- SILC results can be slightly different but the variations remain small.

One remaining issue to resolve can be found in the way pension benefits from the second and third pillar are classified between either public transfers or private transfers/savings/capital income. Currently, pension benefits are treated only as public transfers when they are provided by the state (i.e. the first pillar), while they are considered as private transfer/capital income when provided by the second or third pillar. Consequently, contributions to the second and third pillar are not counted (and deducted) as “direct taxes” (income taxes and social security contributions). This raises an issue as the second-pillar pensions in Switzerland are compulsory and considered part of the public redistribution system in the country. Also, the series provided by Eurostat seemingly includes occupational pensions in “public transfers”. On the other hand, benefits from obligatory and the completely private part of the pension pillar cannot be separated in the data. Furthermore, in other data provisions to the OECD (e.g. Taxing Wages, see Taxing Wages 2009/10, p. 521-522 (www.oecd.org/ctp/taxingwages), the Swiss authorities do not include the contributions paid to the second pillar as social security contributions.

This issue does not impede on the indicators for disposable income but has effects on calculated redistributive effects when comparing per- and post tax/transfer estimates as table 4 shows.

Table 4 Pre- and post tax/transfer poverty rates according to different classification of 2nd pillar pensions

2nd pillar as Threshold = 50% of the current median income (relative poverty) 2nd pillar as transfers capital headcount ratio 7.4% 10.2% Before taxes and transfers median pov gap 48.2% 64.9% headcount ratio 6.7% 6.7% After taxes and transfers median pov gap 21.2% 21.2%

240 5. Summary evaluation

Overall, the OECD time-series for Switzerland match very well with the other available time series. This is particularly relevant when comparing Gini coefficients and poverty rates.

However, the main concern is the lack of time span between the different time-series which prevent from deepening data comparisons. The reasons can be summarised as the following:

 First, the OECD data for Switzerland are only available for three years: 2000, 2004 and 2008;

 Second, the OECD data estimated in 2008 cannot be compared with the 2000-2004 because of the break in the series. General trends over a long time period cannot be drawn due to this break.

 Third, from the period 2000-2004, the OECD is referring to the Income and Consumption Survey handled by the Swiss Federal Statistics Office. The Luxembourg Income Survey is the only survey where data can be found over this period and it is also based on the national Income and Consumption Survey.

 Fourth, the OECD data calculated in 2008 is based on EU-SILC. Therefore, doing comparisons with EU-SILC and Eurostat is somewhat biased given the fact that data on poverty and income inequalities have been only estimated at the European level, in collaboration with the Federal Statistics Office, since 2008 (income year: 2007).

References

Cornali Schweingruber A., Epple R., Oetliker U., Rochat S., “ Une nouvelle méthode pour l’Enquête sur les Revenus et la Consommation (ERC) ”, dans la série “Statistique de la Suisse ”, Office Fédéral de la Statistique, Neuchatel 2007.

Modetta C., Müller B., “ Inégalité des revenus et redistribution par l’Etat - Composition, répartition et redistribution des revenus des ménages privés ”, dans la série “ Statistique de la Suisse”, Office Fédéral de la Statistique, Neuchatel 2012.

OECD (2008), “Growing Unequal? Income Distribution and Poverty in OECD Countries (2008)”, OECD, Paris.

Swiss Statistics Office, Press Release n° 0351-1112-80, « Les ménages avec enfants ont plus de difficultés à faire face à une dépense imprévue », 15-12-2011, in French only.

241 INCOME DISTRIBUTION DATA REVIEW - TURKEY

1. Available data sources used for reporting on income inequality and poverty

1.1 OECD reporting

OECD income distribution and poverty indicators are calculated by the Turkish Statistical Office and come from two surveys:

 Household Income and Consumption Survey (from 1994 to 2005),

 Household Income and Living Conditions Survey (from 2007 onwards).

Data are currently available for 1984, 1994, 2004 and 2007 and 2009.

There was a change in survey weighting in 1994 but the OECD figures between the previous and the new method of calculation yielded very similar results for all indicators, including household income levels. There was a change of survey from 2007 onwards which means that data between 2004 and 2007 are not strictly comparable.

1.2 National Reporting and Reporting in other international agencies:

1.2.1 National reporting:

The Turkish Statistical Institute (TurkStat) started to produce statistics on income distribution with Household Income and Consumption Expenditure Survey in 1987 and carried on an independent survey on income distribution in 1994 and income distribution statistics’ were produced from the household budget survey between 2002 and 2005. Since 2006, TURKSTAT started to conduct an “Income and Living Conditions Survey”.

Other indicators on poverty are available in the poverty statistics section in the Turkish National Statistical Office’s website. However, these data are not included in a yearly publication but are used in regular press releases.

1.2.2 International reporting:

Major poverty indicators were added in the “Income and Living Conditions Survey” from 2006 onwards and were calculated by the National Statistical Office. Before that, TurkStat was producing poverty studies with the World Bank. Since 2006, relative income poverty, which has international comparability, has been calculated based on the results of SILC for Turkey, urban, rural and SR Level 1.

242 Table 37. Characteristics of dataset, Turkey

Household Income and Living Conditions Household Income and Consumption Name Survey Survey Name of the TurkStat TurkStat responsible agency Year Income and Living Conditions Survey has 1987, 1994, 2002, 2003, 2004, 2005 been implemented annually since 2006. and 2009 Data collecting Annually frequency Covered The entire members of the households that population live within the borders of the Republic of Turkey were included within the scope, except the population in aged homes, elderly houses, prisons, military barracks, private hospitals, hotels and child care centers. The immigrant population was also excluded from the scope. Sample size 16 565 households (2011 survey) 13000 households (2009 survey) Sampling method Stratified, multi-staggered, clustered sampling. Rotational design is used in this survey (25% of the sampling size is staying in the sample from one year to the other). Dissemination T+10 months frequency Sampling unit Household Response rate 2009 Figures: the non-response rate is 9 % in 2009. The same rate is 6,6% in urban areas, 3,1% in rural areas. Break in Series Change in survey in 1994 Break in 2004 Data are not available on line before 2004 for some data Websource http://www.turkstat.gov.tr/VeriBilgi.do?al t_id=24

243 2. Comparison of main results derived from sources used from OECD indicators with alternatives sources

2.1. Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among the total population followed a reversed u-shape pattern, between 1984 to 2004 with a spike in 1994. This assumption cannot be compared with national statistics office as historical data are no more available on-line.

From 2004, income inequalities have been decreasing among total population. According to TurkStat, the Gini coefficient is declining from 0.44 in 2002 to 0.38 in 2011. This general trend is confirmed both by OECD database and by TurkStat Office while figures show some discrepancies. In 2004, between OECD and TurkStat, the Gini coefficient is 0.03 higher for OECD (0.43) than for TurkStat (0.40). Interestingly, the Gini coefficient of non-equivalised incomes published by TurkStat matches the OECD reference series in 1994 but is lower in all further years. In later years the Gini coefficient of equalivalised incomes published by TurksStat matches the OECD series.

A longer time series on Gini coefficient is available on TurkStat from 2004 up to 2011 with a breakdown between urban and rural areas. Gini coefficients are slightly higher for urban areas than for rural ones over the estimated period. Please, note that TurkStat is producing Gini coefficients by household disposable incomes and by equivalised household disposable income from 2004 onwards. In the latest press release on Income and Living Conditions Survey (published on 17/09/2012), TurkStat is communicating on the Gini coefficients by equivalised household disposable incomes.

Figure 68. Trends in Gini coefficients (at disposal income), Turkey (1984 – 2011)

OECD reference series

TURKSTAT - Gini coefficient by household disposable incomes - TOTAL

TURKSTAT - Gini coefficient by equivalised household disposable income - TOTAL 0.55

0.50

0.45

0.40

0.35

0.30

1986 1989 1990 1992 1993 1994 1995 1998 1999 2001 2002 2004 2005 2008 2009 2011 1984 1985 1987 1988 1991 1996 1997 2000 2003 2006 2007 2010 1994 old 1994

244 Figure 69. Trends in Gini coefficients (at disposal income; 2004 – 2011)

OECD reference series TURKSTAT - Gini coefficient by household disposable incomes - TOTAL TURKSTAT - Gini coefficient by household disposable incomes - Urban TURKSTAT - Gini coefficient by household disposable incomes - Rural TURKSTAT - Gini coefficient by equivalised household disposable income - TOTAL 0.45

0.40

0.35

0.30 2004 2005 2006 2007 2008 2009 2010 2011

In addition, income quintile share ratios (S80/S20) are confirming the same trends pointed out in the above section. The data are available for 1984, 1994, 2004 and 2007 in the OECD reference survey. In the Income and Living Conditions Survey from TurkStat, data are available from 1994 with a long-time series from 2004 to 2011 with a breakdown by rural and urban areas.

According to the OECD reference survey, the income quintile share ratio remained unchanged from 1984 up to 2004 with an upward spike in 1994. OECD and TurkStat are presenting an identical figure for 2004. From 2004 onwards, the S80/S20 ratio is steadily declining from 9.13 in 2004 down to 8.14 in 2007. The TurkStat figures are more volatile but we can identify a similar tend with declining figures from 2004 onwards. It may be important to note that TurkStat figures on income quintile share ratio are lower than OECD figures in 2004 (9.12 versus 7.7) while being similar in 2007 (8.1 in both surveys). For 2010 and 2011, TurStat’s latest figures on S80/S20 indicators are estimated at 8.00 for both years.

In terms of geographical allocation, income quintile share ratios are slightly higher for urban areas than for rural ones over the estimated period (2004 – 2011).

245 Figure 70. Trends in S80/S20, Turkey (1984 – 2011)

2.1.2 Time series of poverty rates, poverty composition

According to the OECD income distribution database, the share of the Turkish population living with less than 50% or 60% of the median equivalised income has remained stable from 1994 to 2007. Over this period, around 16%-17% of the population was living in relative poverty with a poverty line of 50% whereas this figure was estimated around 24% with a poverty line of 60%.

The national statistics office calculates two-times series of poverty rates:

 One poverty rate calculated by relative poverty thresholds based on income (Turkey) with a poverty line at 40%, 50%, 60% or 70%.

 One poverty rate by relative poverty thresholds calculated for Turkey and based on income (Turkey) with a poverty line at 40%, 50%, 60% or 70%.

Both of these time-series are calculated for urban and rural areas.

In 2007, the only comparable year between the OECD reference survey and the TurkStat databases, figures are similar to the TurkStat poverty rates which are calculated by relative poverty thresholds based on income (Turkey).

The latest figures on poverty are published in a TurkStat press release dating back from 17th September 2012. Figures report that: “16, 1% of total population is at-risk-of poverty according to poverty threshold calculated by 50% of equivalised household disposal median income31”. This rate is estimated at 13.9% for urban areas and at 15.7% for rural ones by using poverty thresholds calculated separately for urban and rural areas.

31 TurkStat, Press release published on 17/09/2012, http://www.turkstat.gov.tr/PreHaberBultenleri.do?id=10902

246 Figure 71. Trends in Poverty rates, after taxes and transfers, Turkey (1994 – 2011)

OECD - After taxes and transfers - Pov Line of 50% OECD - After taxes and transfers - Pov Line of 60% TURKSTAT - Pov rate by relative poverty thresholds based on income (Turkey) - Pov Line of 50% TURKSTAT - Pov rate by relative poverty thresholds (calculated for Turkey) based on income (Turkey) - Pov Line of 50% TURKSTAT - Pov rate by relative poverty thresholds based on income (Turkey) - Pov Line of 60% TURKSTAT - Pov rate by relative poverty thresholds (calculated for Turkey) based on income (Turkey) - Pov Line of 60% 30%

25%

20%

15%

10% 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011

There is no information readily available on child poverty for Turkey at the Turkish Statistical Office.

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for the latest available year, according to the OECD benchmark series. According to the Turkish Statistical Office, the share of salaries and wages in total income is more than other types of income as per 2004 and 2005. Looking at the income shares, 39.2% of annual disposable income of individuals comes from salaries and wages; 28.8% from self-employment incomes; 23% from transfer incomes and 5.6% from property income, interests and dividends. Similar ratios were noticed in 2004 with the following respective figures: 38.7%, 31.8%, 21.2% and 4.9%. For the OECD reference survey, the conclusion is very similar, though the OECD data suggest a lower transfers hare and a higher share of self-employment income

However, the below results are difficult to compare with other countries as taxes are not provided in both cases. Indeed, Turkey is one of the few OECD countries which report all incomes net of taxes.

247 Table 2. Shares of income components in total disposable income, OECD reference series

Figure 6 compares the trend in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in GDP, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series rather different trends throughout the period. The OECD series based on household incomes is recording a significant increase of public transfers throughout the period whereas the OECD Social Expenditure database (OECD SOCX) remained rather stable over the period.

Figure 6 Trends in shares of public social transfers

4. Metadata of data sources which should explain differences and inconsistencies

Differences and inconsistencies are only relevant when dealing with the comparison of the main aggregates of income components. In this category, differences may appear between the OECD reference survey and the Turkish Statistical Office reference survey, with the OECD reference data suggesting a lower share of public transfers in disposable income and a higher share of self-employment income. Slight differences regarding the means and the breakdown of the income components may explain the spread between the different data. For instance, the variable “daily wages” is a component of the disposal income for Turkstat which is not the case for the OECD reference survey.

5. Summary evaluation

Generally speaking, the OECD reference series match with the Turkish series over the comparable time period. Figures related to the income distribution are broadly similar between the two time-series. The Gini coefficients are slightly higher for OECD than for TurkStat.

248 The break in 1994 does not really affect the comparison for two main reasons. First, data are similar or rather close between the previous OECD reference survey (called 1994 old) and the new OECD reference survey (called 1994). Second, the 1994 data from the Household Income and Consumption Survey (before 1994) are usually not available anymore on the Turkish National Statistical Office’s website.

However, the major difficulty lies in the fact that the common period of reference between the different surveys is very limited. The OECD has very limited figures beyond 2004 and the TurkStat has scarce information on-line for figures before 2004. More precisely, for income distribution indicators, the time-series of comparison are starting in 2004 and for poverty indicators, they are starting only in 2007 (OCED data are available before 2007 but not for TurkStat). Therefore, comparisons are difficult to assess for these reference series.

Finally, Turkey is one of the few OECD countries which report all incomes net of taxes which disallows analysis of the redistributive impact of taxes and benefits.

249 INCOME DISTRIBUTION DATA REVIEW–UNITED KINGDOM

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

OECD Income distribution and poverty indicators for the United Kingdom are computed by the UK Department for Work and Pensions (DWP) from the annual Family Resources Survey (FRS). FRS data are available from 2000-01 in the OECD database, from 1992 at DWP. OECD data for 1975, 1985, 1990, 1994-95 (and 2000-01) are from the FRS predecessor: the Family Expenditure Survey (FES). To obtain a consistent series over the whole period, FES data prior to around 2000 have been interpolated for the current FRS definition.

1.2. National reporting and reporting in other international agencies:

Income distribution and poverty indicators for the United Kingdom are also available from:

 ONS’s Living Costs and Food survey (LCF), formally known as the Expenditure and Food Survey (EFS). This series is published in “The Effects of Taxes and Benefits on Household Income” (ETB)

 DWP’s Household Below Average Income based on an adapted version of the FRS and BHPS

 Eurostat’s EU-SILC annual survey since 2005, from 1995 to 2001 from Eurostat’s ECHP annual survey. Between the end of ECHP and start of EU-SILC (2002, 2003 and 2004), data was provided by ONS (probably from FRS).

 LIS database, using the FRS in 1999 and 2004, and FES in 1969, 1974, 1979, 1986, 1991, 1994 and 1995

The below table 1 presents the main characteristics of those four datasets:

250 Table 1. Characteristics of datasets, United Kingdom

2. Comparison of main results derived from sources used for OECD indicators with alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among total population has risen faster in the United Kingdom than in any other OECD country between 1975 and 2000. From a peak

251 in 2000 (0.35) and subsequent fall at 0.33 in 2004, it has been rising again since 2005 at 0.34, above the OECD average of 0.31.

The four other series of Gini coefficients on disposable income in the United Kingdom generally show similar levels and trends since the mid90s. The LIS series (also coming from FES & FRS) shows slightly higher levels of inequality throughout, so does the DWP series particularly from 2005 to 2009. The ONS, the DWP and the EU-SILC series fluctuate around the OECD series across time.

Note that Eurostat identifies a break between the 1999 data point at 0.32 and 0.35 in 2000, so the 1999 data point does not seem comparable with figures from 2000 onwards for this series.

Figure 1.1 Trends in Gini coefficient (disposable income)

Also, when comparing the income quintile share ratio (S80/S20) from the OECD reference survey and the Eurostat EU-SILC since 2000, as for the Gini series, the EU-SILC series from Eurostat fluctuates around the OECD series at around 5.5. Also again, the DWP-HBAI series follows the same trend - but at much higher levels - up to 2005 where there is a sharp increase, before a decrease in 2009 and 2010. The 1- to-2 points difference in S80/S20 between the OECD/Eurostat series and the DWP-HBAI series is still to be determined.

252 Figure 1.2 Trends in income share quintile ratio S80/S20

2.1.2 Time series of poverty rates

According to the OECD income distribution database, the share of the UK population living with less than 50% of the median equivalised income (8 166 £ per year in 2010) decreased sharply from 14% in 1990 to 11% in 1996, and it gradually decreased (with fluctuations) to 10% in 2010.

The OECD reference series show similar trends than other series, similar levels than the Eurostat EU- SILC series, but lower levels than the LIS series and slightly higher levels than the ONS-DWP-HBAI (Households Below Average Income) series.

Figure 2.1 Trends in poverty rates

253 The child poverty series show similar differences than the above poverty rates for total population. The DWP-HBAI child poverty series reported lower rates in the 1990s and 2000s (3 to 5 percentage points), but these differences have reduced to 1 percentage point in 2010.

Figure 2.2 Trends in Child poverty rates

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components shares with alternative data sources

3.1. Comparison of main aggregates: earnings, self-employment income, capital income, transfers and direct taxes

Table 2 shows shares of income components for 2008, according to the OECD benchmark series. When comparing the composition of the average equivalised disposable income of the OECD reference series (based on FRS) with the EU-SILC series, there are differences (between 6-8% of HDI) in shares of capital income (higher in OECD series), transfers and taxes (higher in EU-SILC series).

Table 2. Shares of income components in total disposable income, OECD reference series

254 Figure 3 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

Definitions, methodology, data treatment

The ONS series use a national-specific equivalence scale: 0.67 for first adult, 0.33 for other persons aged 14 and over, 0.20 children under 14 years old. The DWP-HBAI stopped using the McClements equivalence scale in 2004/05 and uses the modified OECD equivalence since. Eurostat uses the OECD modified equivalence scale – whereas the OECD reference series uses the square root of household size.

The ONS series defines a child as a individual aged under 16 or aged 16-19, single, living with parents and in full-time education –whereas the OECD reference series defines a child as under 18 years old.

5. Summary evaluation

Despite several methodological concepts, the OECD reference series of inequality and income poverty generally shows similar trends in inequality and poverty than both the series reported on a national level and the Eurostat series (except a few data points for the latter).

255 INCOME DISTRIBUTION DATA REVIEW – UNITED STATES

1. Available data sources used for reporting on income inequality and poverty

1.1. OECD reporting:

The OECD income distribution and poverty indicators for the United States are provided by US Census Bureau of Statistics and calculated on the basis of internal files from the Annual Socio-Economic Supplement (March Supplement) of the Current Population Survey (CPS ASEC)32.

The CPS ASEC collects information based on a standard income definition (i.e. gross annual income in the previous calendar year), with results disseminated in September. The Census Bureau integrates these data with model-based estimates of income and payroll taxes paid by workers and of quasi-cash benefits received from public programmes (e.g. food stamps). These estimates are combined to survey data to obtain estimates on an extended income definition, released in November each year. These are the files used by the OECD.

Data currently available at the OECD refer to income earned in 1974, 1984, 1989, 1995, 2000, 2005, 2008 and 2010. The only break in the time-series occurred in 1993, when a move to a computer-assisted data collection and changes in top coding lead to a higher response rate from high-income respondents.33 While it is not possible to correct for the latter factor, Burkhauser et al. (2008) suggest that the application of a consistent imputation method would have (slightly) increased estimates of the Gini coefficients for years before 1993 relative to those base on the ‘unadjusted’ internal data.34

1.2. National reporting

The main source for national reporting on income distribution and poverty is the US Census Bureau CPS ASEC. This collects information based on gross annual income in the previous calendar year and provides data grouped by a range of characteristics, such as family status, race and Hispanic origin, age, nativity, region, and residence. This is the source also used by the Luxembourg Income Study (LIS), which relies on the Public use micro-data from the US Census. This PUF relies on top-coding that changes over time; this implies some inconsistencies in the LIS time series that are not present in the OECD database. The CPS ASEC is the most timely and accurate national source on household income and it is the basis for computing the official poverty estimates. The Census Bureau recommends it as the preferred source for national analysis.

32 The Annual Socio-Economic Supplement (March Supplement) provides data concerning family characteristics, household composition, marital status, education attainment, health insurance coverage, foreign-born population, previous year’s income from all sources, work experience, receipt of noncash benefit, poverty, program participation, and geographic mobility. 33 Analysis of the 1993 inequality statistics suggested that the increase in the maximum amounts that could be reported by respondents accounts for about 1.8 percentage points (about one-third) of the 5.2 percent increase in the mean income of people in top quartile of the distribution from 1992 to 1993. When also considering the contribution of the switch to CAPI, the total impact of changes in survey methods raises to over one-half of the increase. See P. Ryscavage, “A Surge in Growing Income Inequality?,” Monthly Labor Review, August 1995, pp. 51-61. 34 Burkhauser R.V., S. Feng, S. Jenkins and J. Larrimore (2008), “Estimating trends in US income inequality using the current population survey: the importance of controlling for censoring”, NBER Working Paper 14247.

256 The CPS ASEC is also used by the Luxembourg Income Study (LIS), which relies on the public use micro-data CPS-ASEC files. Since October 2011, the LIS data (as reported in Inequality and Poverty Key Figures.) moved from a concept of cash disposable household income to an enlarged concept which includes non-monetary income from labour and from both public and private transfers. The public use CPS-ASEC files are affected by several changes in the top coding applied to the income of the richest households, which leads to some inconsistencies in the reported time-series (which are not present in the OECD series).

The main measure used in national reporting is the “official poverty rate”, based on the Office of Management and Budget’s (OMB) Statistical Policy Directive 14. The first official U.S. poverty estimates were released in 1964, and have undergone minimal changes since then (limited to the upgrading of the poverty threshold for annual CPI inflation (CPI-U).35 Since November 2011, the Census Bureau also reports a “supplemental poverty measure”, based on the specifications drawn by an interagency technical working group and on the recommendations of a 1995 National Academy of Science report. The new measure incorporates additional items such as tax payments and work expenses in the estimates of family resources, and on thresholds derived from the Consumer Expenditure Survey referring to expenditures on basic necessities (food, shelter, clothing, and utilities) adjusted for geographic differences in the cost of housing.36

Beyond the CPS-ASEC, the Census Bureau reports poverty data from several other household surveys and programs. In addition, income inequality and poverty estimates are reported by a number of other agencies based on additional national sources, which are described below.

Additional national sources

The Survey of Income and Program Participation (SIPP) is a panel survey providing detailed information on participation in transfer programs. The survey, which is conducted by the Census Bureau, also provides data on different income sources (wages and salaries, cash benefits from social insurance and welfare programs, returns from property, assets, and holdings), as well as labour force participation and health insurance coverage. Data are reported for individuals, families, and households during the time span covered by each of its panels. The survey started in 1984 (the first interview was conducted in October 1983) and was redesigned in 1996. The duration of each panel ranges from 2 ½ years to 4 years. The U.S. Census Bureau is currently redesigning the survey to reduce respondent burden and attrition, and to deliver data on a more timely basis, while keeping its focus on the same topic areas of the earlier SIPP panels.

The Panel Study of Income Dynamics (PSID), directed by the University of Michigan and funded by the National Science Foundation, is the longest nationally representative household panel survey in the world. PSID data have been collected annually from 1968 to 1996 and biennially since 1997. It gathers

35 The CPI-U is used to update the thresholds of the official poverty measure for changes in the cost of living (as indicated by the Office of Management and Budget’s (OMB) Statistical Policy Directive 14). An alternative price index is the Consumer Price Index Research Series (CPI-U-RS), which presents an estimate of the CPI for all Urban Consumers (CPI-U) that incorporates a range of methodological improvements made over that period. The Census Bureau uses the CPI-U-RS, computed by the U.S. Bureau of Labor Statistics from 1977 through 2011 to adjust for changes in the cost of living in income and earnings over time. 36 The official poverty measure is used by all federal agencies in their statistical work. However, government aid programs do not have to rely on it as eligibility criteria. Many government aid programs use a different poverty measure following the Department of Health and Human Services (HHS) poverty guidelines or variants thereof, and each aid program may define eligibility differently.

257 data on family and individuals income, employment, wealth, expenditures, health, marriage, childbearing, child development, philanthropy, education, and numerous other topics.

The Survey of Consumer Finances conducted by the National Opinion Research Center (NORC) of the University of Chicago and sponsored by the Federal Reserve Board, is a triennial cross-sectional survey of US families. While the main focus is the distribution of wealth, SCF also provides information on income distribution, based on information on the family’s cash income, before taxes, for the full calendar year preceding the survey. Data are available for 1962-1963 and every three years from 1983 to 2010. 37 The survey provides information on families’ balance sheets, pensions, income, and demographic characteristics, and includes information from related surveys of pension providers. The income component considered in SCF includes wages; self-employment and business income; taxable and tax- exempt interest; dividends; realized capital gains; food stamps and other support programs provided by government; pensions and withdrawals from retirement accounts; Social Security; alimony and other support payments; and miscellaneous sources of income for all members of the primary economic unit in the household.

The Congressional Budget Office (CBO) occasionally presents analysis on income distribution based on a combination of income-tax files (from the “Statistics on Income”, a nationally representative sample of individual income tax returns from the Internal Revenue Service, IRS) and CPS ASEC data. Data are available from 1979 to 2009. CBO reports separately on market and disposable income. Market income includes labour income (i.e. cash wages and salaries, including those allocated by employees to 401(k) plans), employers’ health insurance premiums and their share of Social Security, Medicare, and federal unemployment insurance payroll taxes); business income (i.e. net income from businesses and farms operated solely by their owners, partnership income, and income from corporations); capital gains (i.e. profits realized from the sale of assets); capital income (i.e. taxable and tax-exempt interest, dividends paid by corporations, positive rental income, and corporate income taxes). Disposable income is the sum of market and transfer income (i.e. Social Security, unemployment insurance, Supplemental Security Income, Aid to Families with Dependent Children, Temporary Assistance for Needy Families, veterans’ benefits, workers’ compensation, and state and local government assistance programs and the value of in-kind benefits such as food stamps, school lunches and breakfasts, housing assistance, energy assistance, Medicare, Medicaid, and Children’s Health Insurance Program) less taxes paid (i.e. payroll taxes, corporate income taxes, federal excise taxes). The CBO adjusts the household income with the square root of the size of the household as equivalence scale.

Sub-national sources

The U.S. Census Bureau conducts the American Community Survey (ACS), which provides annual estimates of median household income and poverty by state and other smaller geographic units. Single- year estimates are available for geographic units with populations of 65,000 or more; estimates based on the polling of 3 years of data are available for counties and places with populations of 20,000 or more; and estimates based on the polling of 5 years of data are available for all geographic units, including census tracts and block groups. Sub-national survey data are available from year 2005 to 2010.

The Census Bureau’s Small Area Income and Poverty Estimates (SAIPE) programme also produces single-year estimates of median household income and poverty for states and counties, as well as population and poverty estimates for school districts. These data are model-based single-year estimates resulting from the combination of data from administrative records, intercensal population estimates, and the decennial census with direct estimates from the American Community Survey (from 2005)38. The

37 Over the periods 1983–1989 and 2007–2009, the survey also collected panel data) 38 Previous years used the CPS ASEC.

258 resulting estimates have lower variances than ACS estimates but are released later. Data are available for years 1989, 1993 and from 1995 to 2010.

The below table presents the main characteristics of the sources above mentioned:

259 Table 1. Characteristics of datasets, United States

National sources Subnational sources International sources Annual Socio-Economic Survey of Income and Program Panel Study of Income Dynamics Survey of Consumer Finances (SCF) Congressional Budget Office American Community Survey(ACS) Small Area Income and LIS database Supplement of the Current Participation(SIPP) (PSID) reports Poverty Estimates (SAIPE) Population Survey (CPS ASEC) program Name Annual Social and Economic Survey of Income and Program Panel Study of Income Dynamics Survey of Consumer Finances (SCF) Two primary sources: the Statistics of American Community Survey (ACS) Small Area Income and Poverty Annual Social and Supplement to the Current Population Participation(SIPP) (PSID) Income (SOI), collected by the Estimates (SAIPE) program Economic Supplement to Survey (CPS ASEC) Internal Revenue Service, and the combines data from the Current Population Current Population Survey (CPS), administrative records, Survey (CPS ASEC) of collected by the Census Bureau. intercensal population estimates, the U.S. Census Bureau and the decennial census with direct estimates from the American Community Survey.

260 National sources Subnational sources International sources Annual Socio-Economic Survey of Income and Program Panel Study of Income Dynamics Survey of Consumer Finances (SCF) Congressional Budget Office American Community Survey(ACS) Small Area Income and LIS database Supplement of the Current Participation(SIPP) (PSID) reports Poverty Estimates (SAIPE) Population Survey (CPS ASEC) program

Name Annual Social and Economic Survey of Income and Program Panel Study of Income Dynamics Survey of Consumer Finances (SCF) Two primary sources: the Statistics of American Community Survey (ACS) Small Area Income and Poverty Annual Social and Supplement to the Current Population Participation(SIPP) (PSID) Income (SOI), collected by the Estimates (SAIPE) program Economic Supplement to Survey (CPS ASEC) Internal Revenue Service, and the combines data from the Current Population Current Population Survey (CPS), administrative records, Survey (CPS ASEC) of collected by the Census Bureau. intercensal population estimates, the U.S. Census Bureau and the decennial census with direct estimates from the American Community Survey.

Name of the responsible U.S. Census Bureau (the CPS is U.S Census Bureau University of Michigan Sponsored by the Board of Governors of the Congressional Budget Office U.S. Census Bureau U.S. Census Bureau LIS cross-national data agency sponsored jointly with the U.S. Federal Reserve System with the cooperation center in Luxembourg Bureau of Labor Statistics (BLS)) of the U.S. Department of the Treasury. Since (having U.S. Census 1992 data collected by the National Opinion Bureau as data provider) Research Center (NORC) of the University of Chicago. Year (survey and Annual income data referring to years From 1983 to 1992 a new panel of 1968-1997 annual data. After 1997 Triennial cross-sectional survey from 1962, but Annual income data referring to years Fully implemented from 2006 (Data 1989, 1993 and from 1995 to 1974, 1979, 1986, 1991, income/wage) from (1959) 1967 to 2010 households was introduced each year in biennual data (starting then from over the 1983–1989 and 2007–2009 periods from 1979 to 2009 referring to year 2005) 2010 data years 1994, 1997, 2000, 2004, February, then 1996 (4-year panel), 1999). the survey collected panel data. 2007, 2010 2001 (3-year panel) , 2004 (2 ½ year ) and 2008 (on-going) panels Period over which income is Calendar year prior to the year in Four-month period preceding the Calendar year prior to the year in Calendar year preceding the survey Calendar year prior to the year in Previous 12 months Previous 12 months from 2005 See CPS ASEC assessed which the data are collected interview month which the data are collected which the data are collected onwards, calendar year prior to the year in which the data are collected before 2005 Covered population Civilian non-institutional population Civilian non-institutionalized Families (Members of the families Households (private and collective) Individuals who did file an income tax Households and from 2006 also Group See ACS Individuals and and it includes military personnel who population living in the United States who moved away from their original return complemented with individual Quarters people (GQ) households live in a household with at least one households are also followed: new covered by the population covered by other civilian adult, regardless of families, members going to an the CPS ASEC whether they institution... as well as new family live off post or on post. All other members, but only until when they Armed Forces are excluded.CPS are part of the family) coverage varies with age, sex, and race. Generally, coverage is larger for females than for males and larger for non-Blacks than for Blacks Private Households. Sample size 75,188 interviewed households The effective sample size can range from 8690 families and 24385 individuals 4422 households (2007) and 3,862 households 295,133 individual tax returns in 2009 2,128,104 household unit and 150,052 for na See CPS ASEC containing 204,983 persons (2011 approximately 14,000 to 36,700 (2009) (2009) statistically matched and group quarters people (2011) survey). interviewed households, depending on complemented with CPS ASEC the panel (43,609 eligible Household records. units in wave 6 of the 2008 Panel) Sample procedure Multistage probability sample for The survey design is a continuous series Panel survey (dynamic longitudinal Dual-frame design (area-probability sample and For SOI: Stratified probability Two-phase, two-stages stratified sampling. na See CPS ASEC CPS, plus an additional sample for of national panels with multistage- follow up of families and their list sample). The area-probability sample samples of tax or information returns; The first-phase sample consists of a ASEC to provide more reliable data on stratified sample. The duration of each descendants originally identified in a provides broad national coverage and a sample For CPS ASEC see the specific definition of two separate samples, Main Hispanic households identified the panel ranges from 2 ½ years to 4 years. combination of three probability of households selected with equal probability. column and Supplemental, each chosen at different previous November samples: a nationally rapresentative The 2007 list sample was selected using a points in time (during the summer preceding sample of families designed by the model applied to a set of statistical records the sample year and in January/February of Survey Research Center at the derived from individual income tax returns by the sample year respectively). The Main University of Michigan (the “SRC the Statistics of Income (SOI) Division of the sample covers approximately 99 percent of sample”), an over-sample of low Internal Revenue Service. The model was used the sample, the Supplementary around 1 income families from the Survey of to rank taxpayers in seven strata ordered by percent. Both the Main and the Economic Opportunity (the “SEO estimated wealth and sample observations with Supplemental samples are chosen in two sample”) and the the 1997 higher levels of predicted wealth at a higher stages. The first stage defines the universe Immigrant refresher sample (to rate. The two samples are then combined with for the second stage and then the second- include individuals who arrived in weights. stage sampling uses 16 sampling strata. the United States after 1968). Then subsequently to second-stage sampling, sample addresses are randomly assigned to one of the twelve months of the sample year. The second-phase sample selection aims to subsample the unmailable and non-responding addresses for the CAPI (Computer Assisted Personal Interview) .

Response rate 83.8% 76.7% (wave 6 of 2008 Panel) 97.4% (core part) and 89.8% 70% (2007), 89% (2009 conditional to 2007) na 97.6% household units and 96.9% group na See CPS ASEC (immigrant part) in 2009 quarters people (2011) Imputation of missing values Yes (hot-deck imputation) Adjustment factors applied for missing Yes Yes na Yes na Not additionally to what rotating groups done by the U.S. Census Bureau Unit for data collection Individuals and households Individuals and households Individuals and families The majority of the questionnaire focus on the Individuals and households (Statistical Individuals and households na Individuals and “primary economic unit” (PEU), which match of each SOI record to a households includes all people in the household who are corresponding economically interdependent with the CPS record on the basis of respondent and/or his or her spouse or partner. demographic characteristics and income. For the households who have not to file tax returns, the remaining CPS records were recorded as households who did not file an income tax return, and their income values were taken directly from the CPS.).

Break in series In 1993 associated to the move from a na See CPS ASEC paper questionnaire to a computer- assisted data collection. The CB doesn't correct for it.

Web source: http://www.census.gov/sipp/overview.h http://psidonline.isr.umich.edu/defa http://www.federalreserve.gov/econresdata/scf/ http://www.cbo.gov/publication/4337 http://www.census.gov/did/www http://www.lisdatacenter http://www.census.gov/cps/ http://www.census.gov/acs/www/ tml ult.aspx scfindex.htm 3 /saipe/ .org/

261

2. Comparison of main results from OECD and alternative sources

2.1 Income

2.1.1 Time series of Gini coefficients and other inequality indicators

According to the OECD income distribution database, income inequality among total population in the United States has steady increased from 1974 to 2010. The rise was at a regular pace until 1993, when the series recorded an upward shift. The 1993 upward shift was followed by a decrease of the Gini index until 1999 and a subsequent rise until 2010, where the Gini index was 0.38. A similar trend emerges from other sources, all based on the Census Bureau’s CPS ASEC.

The national headline Gini index published in the Census Bureau’s CPS ASEC shows a smoother increase in inequality and levels that are between 0.08 and 0.11 points higher than the OECD reference series. The difference can be explained by the use of a different income definition (the Census Bureau does not include taxes nor the value of noncash benefits) and by the fact that the series is computed at the household level, while the OECD reference series refers to distribution among individuals (based on a square root scale applied to household disposable income). The Census Bureau also published (CPS ASEC based) Gini coefficients at the person level, based on a three-parameter equivalence scale and or the experimental post-tax incomes.39 Values of the Gini coefficient based on this definition are lower than the national headlines series (by between 0.05 and 0.09 points), but are still higher than the OECD reference series (by between 0.05 and 0.07).40

The Gini index published by the Congressional Budget Office (CBO) is close to (and identical in some years) to the Census Bureau’s series, but has a more jagged pattern with peaks in 1986 (with a Gini index of 0.433), 1988 (0.41), 2000 (0.452) and 2007 (0.465). On the other hand, the CBO series shows a much clearer decline in the mid-90s and in 2001-2002. The CBO series, being based on income tax records, should better capture developments at the high end of the distribution41.

The LIS time series for the Gini presents the same pattern as the OECD series (both rely in the same equivalence scale) but do not match perfectly due to slightly different disposable income definition and

39 The Census Bureau defines post-tax household income as total household cash income (including realised capital gains) less taxes. They compute post-tax household income both with and without the addition of the earned income tax credit (EITC). 40 When comparing the OECD Gini values before taxes and transfers with the Census Bureau’s equivalised one, the values almost completely coincide, with only small differences due to the different equivalence scale used. 41 Piketty and Saez also found that income concentration began to rise in the late 1970s and continued to grow thereafter. Their analysis is based on published tax return statistics, and it uses a market-income definition. The key advantage of those data, as well as the data used by the CBO, is that they are comprehensive at the top of the income distribution, where much of the change in the income distribution has occurred. One drawback of tax return data alone, however, is that they only cover the portion of the population filing tax returns, so they cannot yield distributional statistics for the full population. In addition, they cannot capture income that is not reported on tax returns. See Piketty and Saez, Income Inequality in the United States.

262

data availability.42 A rise of income concentration from the late 1970s has also been documented by Piketty and Saez and by Burkhauser and others.43

Figure 1.1 Trends in Gini coefficient

OECD Income Distribution Database Census Bureau CPS ASEC (official) Census Bureau CPS ASEC (equivalised) Census Bureau CPS ASEC (experimental post-tax) Congressional Budget Office LIS (CPS ASEC) 0.50

0.45

0.40

0.35

0.30

0.25 Gini (at (at Gini disposableincome)

0.20 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009

The increase of income inequality over time is also corroborated by the income quintile share ratio (S80/S20). The OECD income quintile share ratio increased from 6.9 in 1995 to 7.9 in 2010. An increasing trend, at a higher pace, is also highlighted by the CPS ASEC data on a standard income definition, which do not include taxes and quasi-cash benefits received from public programmes. Data sourced from the Congressional Budget Office, show a rise that is similar to that in the OECD reference series, with differences linked to different data assessments and data availability44. The income quintile share ratio confirms the same pattern of the Gini (and presents the same upward shift in 1993). As for the Gini index, the CPS ASEC series presents higher levels of income inequalities than the OECD reference series, for the same reasons stated above; they are also more variable over time, due to their higher frequency. The CBO series shows a decline in the inequality in mid-90s and in 2001-2002, not shown by other sources, while the decline in 2009, not confirmed by other sources, could reflect revisions after the publication of the CBO report and not incorporated in the report.

42 The LIS series cover 1974, 1979, 1986, 1991, 1994, 1997, 2000, 2004, 2007 and 2010. 43 Richard Burkhauser and his coauthors, using internal Census Bureau data, found that the rate of increase in inequality has slowed substantially since the mid-1990s. They computed Gini indexes using a before-tax, after-transfer measure of household cash income, excluding capital gains, was adjusted for differences in household size using the square root of household size. They found that the Gini index grew at an annual rate of 0.14 percent after 1993, in contrast to a growth rate of 0.74 percent in the 1975–1992 period. Even though they found little increase in income inequality after 1993, their analysis did not reject the possibility that inequality could have increased among the highest income households, so they concluded that their results were not inconsistent with those of Piketty and Saez. An increase among the highest-income households may explain the slower growth in measured income inequality after 1993. (Richard Burkhauser and others, Estimating Trends in US Income Inequality Using the Current Population Survey: The Importance of Controlling for Censoring, Working Paper 14247 (Cambridge, Mass.: National Bureau of Economic Research, August 2008). 44 As the OECD reference series covers nine data points (1974, 1984, 1989, 1995, 2000, 2005, 2008, 2010), while the CBO series is an annual data series from 1979 to 2009.

263

Data from the Survey of Income and Program Participation (SIPP), which allows examining changes in the annual income of the same households, are available for the years 1997, 1999, 2001, 2003, 2004 and 2006. These data confirm the trend of increasing inequality, with a peak in 200345.

Figure 1.2 Trends in income share quintile ratio S80/S20

OECD Income Distribution Database Census Bureau CPS ASEC (official) CB Survey of Income and Program Participation (SIPP) Congressional Budget Office 18

8 S80/S20 6

0 1967 1970 1973 1976 1979 1982 1985 1988 1991 1994 1997 2000 2003 2006 2009

2.1.2 Time series of poverty rates

According to the OECD database, the share of the population living with less than 50% of the median equivalised income (29 056 dollars per year in 2010) has increased continuously since 1974, reaching 17.4% in the 2010, 2 points higher than in 1974. The poverty rate increased from 1979 to 1983, flattening until 1993 (with a peak of 18.4% in 1987) and then slightly declining until the beginning of the 2000s. Since then, the series resumed its upward trend.

A similar pattern, but more pronounced and variable, emerges when looking at the official poverty figures published in the CPS ASEC. Compared to 1974, poverty increased more than 3 point reaching 15% in 2010. Unlike the OECD reference series, the official poverty rate declined from 1983 to 1990 and from 1993 to 2000; it also declined before 1974. The official poverty estimates are between 2 and 5 percentage points lower than the OECD reference series. In the Figure below, both the Census Bureau’s official poverty measure (using the Consumer Price Index for All Urban Consumers, CPI-U) and the measure based on the Consumer Price Index Research Series Using Current Methods (CPI-U-RS) are presented. The two measures differ only for the years before mid-90s, with the CPI-U-RS series showing higher poverty. The Census Bureau’s Supplemental poverty measure is also shown for the years for which it is available (2009 and 2010). It shows a pattern similar to the official poverty, but slightly higher as it incorporates additional items as tax payments.

45 This is explained by the fact that only the top quintile in the panel experienced a significant increase in the share of total household income between 2001 and 2003, while for the other quintiles it remained statistically unchanged. ( Dynamics of Economic Well-Being: Fluctuations in the U.S. Income Distribution, 2004–2007 http://www.census.gov/prod/2011pubs/p70-124.pdf)

264

An average monthly poverty measure and a poverty measure calculated using the annual income and threshold46 are published by the Census Bureau based on the Survey of Income and Program Participation (SIPP) for the years 1987, 1990-98, 2001-07 and 2009 (2001-2006 only for the annual income measure). The two series display patterns and levels similar to the CPS ASEC series, except in mid-90s, where the SIPP monthly average series peaks, while the CPS ASEC series decreases. The annual SIPP series is lower than below the CPS ASEC series SIPP typically reports more income (and therefore lower annual poverty) than the CPS ASEC.

The LIS series has pattern and magnitude similar to the OECD series, with minor differences due to the LIS more limited time coverage. The small differences for the same year can be explained by different definitions of disposable income in the two sources.

Figure 2.1 Trends in poverty rates

OECD Poverty rate (relative - 50% median) CB CPS-ASEC (Official national poverty) CB CPS ASEC (CPI-U-RS) CB Supplemental poverty measure CB SIPP (annual income and threshold) LIS (CPS ASEC) CB SIPP (average monthly)

20%

18%

16%

14%

12%

Povertyrate 10%

6% 1966 1971 1976 1981 1986 1991 1996 2001 2006 2011

2.2 Wages

See Part II of the present Quality Review.

3. Consistency of income components across different surveys

3.1. Comparison of main aggregates: earnings; SE income; capital income; transfers; taxes

When comparing the composition of the average equivalised disposable income of the OECD reference survey (based on CPS ASEC) with the CBO series (based on the tax files and CPS ASEC), shares of disposable income generally match. Wages represent a slightly higher share in the OECD reference series, as the CBO disposable income series includes other components (e.g. capital gains, other income, additional transfers) not included in the CPS ASEC. Furthermore, the CBO series shows income levels that are always higher than in the CPS ASEC, as the income taxes records tend to capture better high-income households. Additionally the tax files collects detailed information on income components (as

46 Panel and yearly estimates contain different samples. 3-year panel estimates include only respondents in the panel for 10 waves whereas calendar year estimates include people in sample for 12 months. The total number of respondents in each sample is as follows: 27,840 in the 3-year panel, 86,128 in 2004, 76,953 in 2005, and 34,372 in 2006. In wave 9 of the SIPP 2004 Panel, there was a 53% sample reduction. However, the calendar year weight for 2006 and the 3-year panel weight correct for that sample reduction.

265

capital gains and other income from capital) excluded from the OECD reference series, as well as on taxes, which are estimated by the Census Bureau in the OECD series.

Table 2. Main income aggregates, USA

Survey Year Unit Wages Capital Self Employment Transfers Taxes Disposable income (HDI) OECD reference suvery CPS ASEC 2008 natcur 35,109 3,766 2,318 3,779 -9,256 35,717 % av HDI 98.3% 10.5% 6.5% 10.6% -25.9%

Income Capital gains Business income Other income Other INCOME survey SOI & CPS ASEC 2008 natcur 55,894 6,512 3,739 5,105 5,785 10,258 -15,761 71,532 % av HDI 78.1% 9.1% 5.2% 7.1% 8.1% 14.3% -22.0% Figure 4 compares trends in shares of public cash transfers in equivalised disposable income from the OECD reference series with the share of total cash social spending in net national income, reported from the OECD Social Expenditure database (OECD SOCX). OECD SOCX series include pensions, incapacity, family, unemployment, social assistance. Both series show similar trends throughout the period. The two series display a similar profile, with trends over time moving in the same direction..

Figure 4 Trends in shares of public social transfers

4. Metadata of data sources which could explain differences and inconsistencies

There are several methodological differences between the OECD Terms of References and the methodology used by the Census Bureau for its own national publications based on the CPS ASEC. The main differences are as follows:

 Unit of report and equivalence scale: Inequality is generally reported by the Census Bureau across households rather than across individuals, and no ‘equivalisation’ is applied to official income data. 47

47 Nevertheless the Census Bureau publishes also selected measures of equivalence-adjusted income dispersion. The equivalence adjustment used is based on a three-parameter scale that reflects that: i) on average, children consume less than adults; ii) as family size increases, expenses do not increase at the same rate; and iii) the increase in expenses is larger for a first child of a single-parent family than for the first child of a two-adult family. Specifically, the scale fixes the ratio of the scale for two adults and one adult to a constant value 1.41. For single parents the scale adds the number of adults to 0.8 for the first child plus 0.5 times all other children raised to a power of 0.7, that is (A + 0.8 + 0.5 * C)^0.7. All other families use the formula (A + 0.5 *C)^0.7.

266

 Income definition: The income concept used by the Census Bureau is that of money income before payments for personal income taxes, social security, union dues, etc. This concept excludes the value of noncash benefits, such as those provided by the Supplemental Nutrition Assistance Program (SNAP), Medicare, Medicaid, public housing, and employer-provided fringe benefits. The OECD income definition also includes items that are imputed by the Census Bureau (e.g. food stamps, income and wealth taxes, EITC);

 Poverty thresholds: The Census Bureau reports poverty based on the OMB official definition, based on a set of absolute thresholds (which do not vary geographically) adjusted by family size, gender of the family head, number of children under 18 years, and farm-nonfarm residences to determine who is in poverty. The official poverty definition uses money income before taxes and tax credits, and excludes capital gains and non-cash benefits. In the OECD database, poverty is defined using relative thresholds expressed as a given percentage (50% and also 60%) of the median equivalised disposable income of the entire population. The OECD also reports on relative poverty anchored in time (mid-1990s and 2005)

Differences among the Census Bureau’s surveys and with the PSID:

 Taxes and assets holdings: While in the CPS ASEC no taxes are included, in the Survey of Income and Program Participation (SIPP) and in the Panel Study of Income Dynamics (PSID) the Census Bureau collect information on some federal, state and local income taxes, payroll taxes and property taxes, as well as on assets holdings (in the CPS ASEC only home ownership is included). In SIPP, a detailed inventory of real and financial assets and liabilities is collected once a year for panels from 1996 (and at least once per panel in prior years); in the PSID regular information about home value and mortgage debt is available, but information about wealth is collected occasionally.

 Definition of the unit of collection: The PSID does not collect information on income of household members who are not members of the PSID family. Also, the CPS definition of household differs from the PSID concept of a family when one or more PSID families reside in the same household.48 For these reasons PSID estimate of family income tend to be lower than the CPS estimate of household income49. The household composition is defined in the interview month (February, March, or April) in the CPS ASEC, while in the SIPP family composition may vary during the reference period. This affects the selection of the poverty threshold; as a result, the annual poverty rate estimates in the SIPP differ from official estimates based on the CPS ASEC.

48 This happens, for example, when a grown child marries and leaves the parental home to live independently, but then eventually comes back to live with their parents. It is PSID practice to treat the parent’s family and the new adult child’s family (even if it consists of a single person) as separate “families” and obtain full, independent interviews from both of them. 49 Results of a study of Elena Gouskova, Patricia Andreski, and Robert F. Schoeni (Comparing Estimates of Family Income in the Panel Study of Income Dynamics and the March Current Population Survey, 1968- 2007), where the authors compared estimates of family income between the PSID and the CPS from 1968 through 2007 survey years using visualization techniques to assess qualitatively the disparities in the empirical distributions of income, show that the distributions match fairly closely in the range between the 5th and 95th percentiles: historically the PSID estimates have been somewhat higher than the CPS estimates, but the trends are quite similar. The two data sets show less agreement at the upper and lower five percentiles of the distribution.

267

Differences between the OECD reference series and the CBO reported data

While the OECD reference series is based on the CPS ASEC, the income-tax records from the “Statistics on Income” represent the core part of the CBO’s report. These records are then statistically matched to the corresponding CPS records. When comparing the two sources, CBO’s series result in higher income levels and higher inequality estimates than those in the OECD database, due to better and more detailed coverage of high-incomes households, the inclusion of capital gains and a better coverage of other income from capital.

5. Summary evaluation

Differences in methodology and definitions among the sources lead to differences in levels in all the indicators considered. However, all series display similar trend patterns, except for the poverty rates where concepts (‘absolute’ poverty, in the official definition; relative poverty, in the OECD ones) differ. The limited time-coverage of the OECD series does not allow capturing all the variability of the national series (this is the case for the OECD reference series on the income quintile share ratio in respect to the Census Bureau’s series). As taxes and some other income components are estimated in the OECD reference series by the Census Bureau, it could be useful to consider the corresponding values collected in the SOI in the estimation process or compare the estimates to them ex-post. Attention should also be paid when comparing the OECD series before and after 1993, as the move from a paper questionnaire to a computer- assisted data collection in 1993 (for which the Census Bureau does not correct) resulted in a better measure of the income of the wealthier households, and slightly higher measure of income inequality in that year.

268