The Consistency of Cross-sectional and Longitudinal Data in EU-SILC Countries when Measuring Income Levels, Inequality, and Mobility

Transcription

1 The Consistency of Cross-sectional and Longitudinal Data in EU-LC Countries when Measuring Income Levels, Inequality, and Mobility Joachim R. Frick & Kristina Krell (, DIW Berlin) 2nd EU-Microdata User-conference, Mannheim, March 31 - April 1, 2011

2 Motivation Explore problematic features of EU-LC income data and find suggestions for further quality research Comparison of data for Germany with alternative German household survey () shows the extensive influence of survey methods on income measures (Frick & Krell 2010) Unexpected developments of income levels, income mobility, and inequality in some countries Inconsistencies in cross-sectional and longitudinal data 2

3 Structure of this presentation 1. Basics of EU-LC (and ) 2. Definitions 3. Comparative Analyses Income levels, Inequality, Poverty Cross-sectional vs. longitudinal datasets Income & Poverty Mobility 4. Conclusion & recommendations 3

4 1. Basics of EU-LC (and ) Types of interview: CAPI, PAPI, CI, self-administred, proxy-interview Cross-sectional and 4-wave longitudinal sample in most countries Rotational Panel design: 25% of households replaced each year German : started 1984 including ~ households with persons; pure panel, personal interviews, no proxy-interviews ( Country groups: Former Transition countries: Poland, Czech Republic, Slovakia, Slovenia, Hungary, Estonia, Latvia, Lithuania Mediterranean countries: Spain, Portugal, Italy, Greece, Cyprus Continental and liberal countries: Austria, France, Netherlands, Belgium, Luxembourg, Germany, United Kingdom, Ireland Scandinavian countries: Sweden, Norway, Finland, Denmark, Iceland 4

5 2. Definitions Population: Individuals in private households Observation years and periods: cross-sectional: , longitudinal: 2005/2006, 2006/2007, 2007/2008 (income reference period usually observation year-1) Income: Equivalent disposable household income Sum of gross income components of all household members (income from employment and capital, private and public transfers, private and public pension) minus direct income taxes and social security contributions (without Imputed Rent) using the modified OECD-scale (1; 0.5; 0.3) Income in prices of 2005 Imputation of missing income components 5

6 3. Comparative Analyses Cross-sectional perspective: Income Levels, Inequality, Poverty Cross-sectional vs. longitudinal datasets Longitudinal perspective: Income & Poverty Mobility 6

7 Income levels 1,80 Transition Countries 1,80 Mediterranean Countries 1,60 1,60 GR 1,40 1,40 1,80 Continental Countries 1,80 Scandinavian Countries 1,60 1,40 DE-LC 1,60 1,40 IS 7

8 Income inequality Gini coefficient Transition countries Mediterranean countries GR Continental countries Scandinavian countries DE-LC IS 8

9 Relative Income Poverty rate* 1,40 1,30 Transition Countries 1,40 1,30 Mediterranean Countries GR 0,70 0,70 1,40 1,30 Continental Countries DE-LC 1,40 1,30 Scandinavian Countries IS 0,70 0,70 * Poverty threshold given at 60% of Median National Equivalent Income. 9

10 3. Comparative Analyses Cross-sectional perspective: Income Levels, Inequality, Poverty Cross-sectional vs. longitudinal datasets Longitudinal perspective: Income & Poverty Mobility 10

11 Indicator: fraction /CS 1,15 1,05 0,95 0,85 Mean income (Population 2005/2006) 2006 (Population 2005/2006) 2006 (Population 2006/2007) 2007 (Population 2006/2007) 2007 (Population 2007/2008) 2008 (Population 2007/2008) GR DE SOE EP I IS I 1,15 FGT(0) ,05 0,95 0,85 GR DE IS 11

12 1,15 Gini-Coefficient ,05 0,95 0,85 GR DE IS MLD ,70 0, (Population 2005/2006) 2006 (Population 2005/2006) 2006 (Population 2006/2007) 2007 (Population 2006/2007) 2007 (Population 2007/2008) 2008 (Population 2007/2008) 0,50 0,40 GR DE IS 12

13 Why do cross-sectional and longitudinal results differ? Balanced Panel Design causes underrepresentation of poorer population subgroups (deceased, births, migrants) Attrition bias (extremes more likely to leave the panel) Non-inclusion of households with Partial Unit Non-Response (,,,,, + register countries): maybe biased (Verma & Betti 2010) No sufficient correction of attrition or achievement of representetiveness via weighting factors in some countries? 13

14 3. Comparative Analyses Cross-sectional perspective: Income Levels, Inequality, Poverty Cross-sectional vs. longitudinal datasets Longitudinal perspective: Income & Poverty Mobility 14

15 Income Mobility: Comparison of longitudinal populations 2005/2006, 2006/2007, and 2007/2008 Matrix mobility: Normalised Average Jump Matrix mobility: Normalised Average Jump 0,31 0,31 ion 2006/2007 Mobility of Populati 0,29 0,27 0,25 0,23 0,21 0,19 0,17 0,15 0,13 GR IS 0,13 0,15 0,17 0,19 0,21 0,23 0,25 0,27 0,29 0,31 Mobility of Popula ation 2007/2008 0,29 0,27 0,25 0,23 0,21 0,19 0,17 0,15 0,13 0,13 0,15 0,17 0,19 0,21 0,23 0,25 0,27 0,29 0,31 Mobility of Population 2005/2006 Mobility of Population 2006/

16 Income Mobility: Comparison of longitudinal populations 2005/2006, 2006/2007, and 2007/2008 Shorrocks Equalisation Measure Shorrocks Equalisation Measure Mobility of Populatio on 2006/2007 0,22 0,2 0,18 0,16 0,14 0,12 0,1 0,08 0,06 0,04 IS GR 0,04 0,06 0,08 0,1 0,12 0,14 0,16 0,18 0,2 0,22 Mobility of Populatio on 2007/2008 0,22 0,2 0,18 0,16 0,14 0,12 0,1 0,08 0,06 0,04 0,04 0,06 0,08 0,1 0,12 0,14 0,16 0,18 0,2 0,22 Mobility of Population 2005/2006 Mobility of Population 2006/

17 Poverty Mobility: Comparison of longitudinal populations 2005/2006, 2006/2007, and 2007/2008 Mobility of Popu ulation 2006/ ,0 15,0 13,0 11,0 9,0 7,0 5,0 Poverty Mobility: share of mobile population GR IS 5,0 7,0 9,0 11,0 13,0 15,0 17,0 Mobility of Popu ulation 2007/ ,0 15,0 13,0 11,0 9,0 7,0 5,0 Poverty Mobility: share of mobile population 5,0 7,0 9,0 11,0 13,0 15,0 17,0 Mobility of Population 2005/2006 Mobility of Population 2006/

18 Factors influencing mobility Type of interview: personal interview, self-administred by respondent (no interviewer assistance), telephone interview, proxy interview Imputation method: use of cross-sectional or cross-sectional and longitudinal information (Rässler & Riphahn 2006, Starick & Watson 2009, Spieß & Goebel 2005) Rotational panel design: income of recent surveys often more volatile due to the lack of experience in answering income questions (Frick et al. 2006) 18

19 Comparison of income, poverty, and inequality in two longitudinal populations for one year 1, Mean Median FGT(0) Gini 0,70 19 GR DE IS 0,70 GR DE IS 2007 Mean Median FGT(0) Gini

20 4. Conclusion & Recommendations Overall: Large variance in development of income levels, inequality, and mobility Countries with inconspicuous values in cross-sectional data can be conspicuous in the comparison of cross-sectional and longitudinal data All such surveys have to deal with difficulties in measuring economic outcomes (eg. Income) due to non-response, measurement error (self-administration, proxy interviews), Imputation of missing data is based on assumptions and normative decisions and can have great effects on mobility 20

21 4. Conclusion & Recommendations Further check of data quality (esp. weighting?) in countries with problematic measures of income inequality and mobility in cross-sectional and longitudinal datasets robustness check via alternative surveys Better, transnationally and intertemporarily comparable documentation of imputation methods and extent of imputation Use of longitudinal data for imputation in all countries Improvement of infrastructure for substantive and survey methodology research (provision of link between crosssectional and longitudinal data) Continuous ex-post quality check (incl. eventual revisions for all waves to ensure consistency across time) 21

22 References Frick, J. R., Goebel, J., Schechtmann, E., Wagner, G. G., Yitzhaki, S. (2006): Using Analysis of Gini (ANoGi) for detecting whether two sub-samples represent the same universe: The German So- cio-economic Panel Study () Experience, in: Sociological Methods & Research, 34 (4) pp Frick, J.R. and Krell, K. (2010): Measuring Income in Household Panel Surveys for Germany: A Comparison of EU-LC and. paper 265, Berlin: DIW. Lelkes, O., Medgyesi, M., Tóth, I. and Ward, T. (2009): Income distribution and the risk of poverty. In: Ward et al (2009), Chapter 1. Rässler, S., Riphahn, R. T. (2006): Survey Item Non-Response and its Treatment, in: Allgemeines Statistisches Archiv, 90, pp Starick, R., Watson, N. (2007): Evaluation of Alternative Income Imputation Methods for the HILDA Survey. HILDA Project Technical Paper Series No. 1/07, Melbourne Institute of Applied Economic and Social Research, University of Melbourne. Spieß, M., Goebel, J. (2005): On the Effect of Item Nonresponse on the Estimation of a Two-Panel-Waves Wage Equation, in: Allgemeines Statistisches Archiv, 1, pp Verma, V. and Betti, G. (2010): Data accuracy in EU-LC. In: Atkinson, A. and Marlier, E. (2010): Income and living conditions in Europe. Eurostat Statistical Book. 22

23 Thank you for your attention. Kristina Krell -homepage 23

24 Mean income Mean income GR DE IS Country

25 0,450 0,400 0,350 0,300 0,250 0,200 0,150 0,100 0,050 0,000 Inequality Gini-Coefficient Gini Coefficient GR DE IS

26 0,300 0,250 0,200 0,150 0,100 0,050 0,000 Poverty rate FGT(0) FGT(0) GR DE IS

27 Development of median income Transition countries Mediterranean countries 1,70 1,70 1,50 1,50 GR 1,30 1,30 Continental countries Scandinavian countries 1,70 1,70 1,50 DE-LC 1,50 IS 1,30 1,30 27

28 Income inequality Mean Log Deviation 1,50 1,40 1,30 Transition countries 1,50 1,40 1,30 Mediterranean countries GR 0,70 0,70 0,60 0,60 1,50 Continental countries 1,50 Scandinavian countries 1,40 1,30 DE 1,40 1,30 IS 0,70 0,70 0,60 0,60 28

29 Shorrocks equalisation measure 0,19 0,21 0,23 Population 2005/2006 0,19 0,21 0,23 Population 2006/2007 0,19 0,21 0,23 Population 2007/ GR IS DE 0,05 0,07 0,09 0,11 0,13 0,15 0,17 0,19 GR IS 0,05 0,07 0,09 0,11 0,13 0,15 0,17 0,19 0,05 0,07 0,09 0,11 0,13 0,15 0,17 0,19

30 Matrix-mobility: Normalised Average Jump DE 0,27 0,29 0,31 Population 2005/2006 0,27 0,29 0,31 Population 2006/2007 0,27 0,29 0,31 Population 2007/ GR IS 0,13 0,15 0,17 0,19 0,21 0,23 0,25 GR IS 0,13 0,15 0,17 0,19 0,21 0,23 0,25 0,13 0,15 0,17 0,19 0,21 0,23 0,25