Advanced Methodology for European Laeken Indicators

Transcription

1 Advanced Methodology for European Laeken Indicators New Techniques and Technologies for Statistics Advanced Methodology for European Laeken Indicators Brussels, 23. February 2011 University of Trier, Faculty IV, VWL Economic and Social Statistics Brussels, Münich/Kolb/Zins 1 (18) AMELI

2 Outline An Overview of the AMELI Project Brussels, Münich/Kolb/Zins 2 (18) AMELI

3 Advanced Methodology for European Laeken Indicators Project is funded by the European Commission within the seventh Framework Programme Social Sciences and Humanities Area 6.2 Developing better indicators for policy DG RTD in cooperation with DG ESTAT Project ocer: Dr. Ian Perry EC contribution Me Co-ordinator: Ralf Münnich Homepage: Brussels, Münich/Kolb/Zins 3 (18) AMELI

4 Research Interest of AMELI Laeken indicators (partially as prototype) Investigation of EU-SILC datasets from selected countries Improving the statistical methodology Consideration of the entire indicator process Delivery of open source codes under R Investigation of methodology within a simulation framework Adequate graphical presentation of indicators Partners: University of Trier, German Federal Statistical Oce, University of Applied Sciences Northwestern Switzerland, Swiss Federal Statistical Oce, Statistics Austria, Statistics Finland, University of Helsinki, Vienna University of Technology, Statistical Oce of the Republic of Slovenia, and Statistics Estonia Partner Project: SAMPLE (Co-ordinator: Monica Pratesi) Brussels, Münich/Kolb/Zins 4 (18) AMELI

5 Spiderplot of Six Indicators R4 R3 R5 R2 R6 R1 R7 R11 R8 R10 R9 Brussels, Gini Münich/Kolb/Zins QSR Arpr Share of 5 (18) persons with AMELI low educational attainment

6 Motivation of the Work on Composite Indicators Two main reasons why composite indicators (CI) are useful for measuring poverty 1. General advantage of CI: A single number is easier to handle than a large set of indicators 2. Specic advantage in the eld of poverty measurement: Poverty is multidimensional, so a single indicator does not represent the whole range of the phenomenon (Materazzo/Pareto 2009 ) Main problem: It is not a simple task to construct a CI The nal outcome has to be interpreted carefully Application of sensitivity analysis (JRC group of Andrea Saltelli) Brussels, Münich/Kolb/Zins 6 (18) AMELI

7 Denition 1 Let y t i,c Composite Indicators be the outcome of a single indicator i (i = 1,..., ν, with ν variables of interest), country c (c = 1,..., C ) and year t (t = 1,..., T ). A composite indicator is a function Ψ c,t = Ψ c,t ( ) y1,c, t y2,c, t..., y t ν,c : R ν R 1 Goal of composite indicators Benchmarking To build ranks between countries (of interest) Ψ c1,t Ψ c2,t Development To view indicator development over time Ψ c,t Ψ c,t 1 bzw. T 1 Ψ c,t /Ψ c,1 1 (mean growth rate) Brussels, Münich/Kolb/Zins 7 (18) AMELI

8 CI Construction Scheme Sample Selection Normalisation Delete an indicator or not Weighting aggregation Composite Indicator Brussels, Münich/Kolb/Zins 8 (18) AMELI

9 Variance Decomposition Main idea behind the method is to decompose the function f ( ) into terms which only depend on 1, 2,..., k input triggers By calculating the variances of dierences of those terms you receive the impact of the dierent input triggers on the variance of output First-order sensitivity index of the nth input trigger: S n = V [E(Y I n)] V (Y) Possible expansion to higher order eects (Saltelli et al. 2008) In this study we only used the rst order and total order eects Brussels, Münich/Kolb/Zins 9 (18) AMELI

10 Boxplot: All Composite Indicators R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R Brussels, Münich/Kolb/Zins 10 (18) AMELI

11 Boxplot: Only Z-Scores Used R1 R2 R3 R4 R5 R6 R7 R8 R9 R10 R11 Brussels, Münich/Kolb/Zins 11 (18) AMELI

12 Measurement of Development To monitor the process towards agreed policy goals we are interested in the evolution of social indicators. EU25 EU15 EA ARPR GINI R Source: Eurostat, 2009, Reading naively point estimator tables may lead to over-interpret the data Was the change (in time) of an indicator value signicant or not? How to measure signicant changes of ARPR, GINI, and QSR? Brussels, Münich/Kolb/Zins 12 (18) AMELI

13 Two problems arise: 1. The statistics in question (the Laeken indicators) are highly non-linear. Basic variance estimation formulas cannot be applied directly. 2. The Surveys used to estimate the indicator values (EU-SILC) are often time dependent. The correlation through time between indicators has to be taken into account. Covariance estimation for non-linear statistics Ĉov(Ît0, Ît1) ( = Ĉov i S t 0 j S t 1 u j ) π j u i, π i i S t 0 j S t 1 = ( 1 π i π j π ij π i = P(i S t.); π ij = P(i S t0, j S t1 ) Brussels, Münich/Kolb/Zins 13 (18) AMELI ) ui π i uj π j

14 Test of Signicant Change: H 0 : = 0 = ARPR RMPG GINI QSR Hulliger (2005), Displays of indicators and of their accuracy, Conference on Visualizing and Presenting Indicator Systems Brussels, Münich/Kolb/Zins 14 (18) AMELI

15 Evidence Based Policy Marking and Quantitative Policy Goals At the Lisbon Council of March 2000, social cohesion was dened as one of the most challenging responsibilities for the European Union (cf. Atkinson et al. 2005, p. 17). To monitor the progress towards agreed policy targets, a rst set of social inclusion indicators was nally decided on at the Laeken Council in December These indicators should play a crucial role in the Open Method of Coordination (OMC) on Social Inclusion and are considered as a reliable source policy makers can base their decisions on. For the coming decade a new agenda was set up, Europe In the eld of Poverty and social exclusion the target was set to have least 20 million fewer people in or at risk of poverty and social exclusion by Beil, Kolb, and Münnich (2011) Brussels, Münich/Kolb/Zins 15 (18) AMELI

16 Laeken Indicators in Ocial Documents To uncover the use of the Laeken primary indicators a document analysis of the NSRP's was carried out. The resulting pattern can be seen in the following gures. Both gures show the absolute number of primary indicators used in the main text of the report and the number of indicators that was used only in the annex of the report. Brussels, Münich/Kolb/Zins 16 (18) AMELI

17 Results from the Online Survey The indicators SIP1, SIP4 and 5 are most known and most in use. Child well being (SIP11) is a relatively new indicator which has to be established. Most respondents consider the Laeken Indicators as an appropriate instrument to measure poverty. 36 % of the respondents did not know the Laeken Indicators - confusion exists because of changing name. Following the opinions reected in the questionnaire, the Laeken portfolio comprehends an appropriate number of indicators and the indicators are balanced across all dimensions. Brussels, Münich/Kolb/Zins 17 (18) AMELI

18 Summary and Outlook The project will be nalized on 31st March 2011 The nal outcome will be available after acceptance of the deliverables The AMELIA (synthetic) dataset will be made available for further comparative research Some important aspects of AMELI... Risto Lehtonen Estimation of poverty indicators for domains Monique Graf Information hidden in the Laeken monetary indicators Beat Hulliger Main results of the AMELI simulation study on advanced methods for Laeken indicators Matthias Templ Visualization of indicators within the AMELI project Brussels, Münich/Kolb/Zins 18 (18) AMELI