Regional Data from Sample Surveys using the lfs as an example

Transcription

1 Regional Data from Sample Surveys using the lfs as an example Susanne Schnorr-Bäcker Federal Statistical Office of Germany Klaus Trutzel National Urban Audit Coordinator SCORUS Conference Brussels 4-5 June 2013

2 Agenda 1. Basics 2. Data for Urban Audit from sample surveys: The German way 3. Further developments 4. Conclusions Folie 2

3 1. Basics Regional Structure of Germany* NUTS 0 Deutschland 1 NUTS 1 Länder/federal states 16 NUTS 2 Administrative regions 19 NUTS 3 Kreise/administrative districts 402 LAU 1 Associations of municipalities LAU 2 Municipalities Cities > inhabitants 14 Cities > inhabitants 39 Cities > inhabitants 80 Cities > inhabitants 189 *) Regional units: Gemeindeverzeichnis, December 31st, 2012 Folie 3

4 Urban Audit Regional units: 84 cities at NUTS 3* Data requirements for households and individuals Such as household structure net income living conditions education employment commuting mostly from Microcensus (= Labour Force Survey for Germany) *since UA 2010: totally 125 cities with 88 cities at NUTS 3 Folie 4

5 Micro-Census (MC) Characteristics biggest sample survey in the European Union covers 1 % of households per year reduces burden by 25 % rotation per year Sample Design: stratified random sample at regional level for different housing structures Folie 5

6 Sample Design for Micro-Census 2. Buildings: 4 strata amongst 1-4 dwellings 5-10 dwellings > 11 dwellings 1. Regional strata at NUTS 3 level Every city > inhabitants (i.e inhabitants) Every Landkreis > (i.e inhabitants) 350 exact NUTS 3 (of 389 strata in total) Folie 6

7 Extrapolation of Micro-Census 1. Standard extrapolation for NUTS 0 Germany at NUTS 2 level Extrapolation by external sources, in particular Population register Register of foreign population Registers for other groups of people such as in defense and in civil service Sample results for 130 regional adjustment areas (RAS) 2. Regionalized extrapolation Linear regression with bounding parameters according to Mikro-Census results Population register at regional substrata-level Folie 7

8 Disaggregating the German Micro-Census by combining it with regional statistics on the level of NUTS3 Using data from: Micro-Census Regional database Regionaldatenbank Deutschland Federal Employment Agency Folie 8

9 2. Data for Urban Audit from sample surveys: The German way The German Micro-Census is a 1-percent sample of private households. Each household is contacted 4 times within a 4-year period and then replaced. The lowest territorial level for which results are publicly available are the regional adjustment strata (RAS) = groups of NUTS3 units. Until reference year 2004, results for RAS were the only Micro-Census data source available for estimations. In 2009, an agreement with the Federal and the State Statistical Offices was reached to use Micro-Census data on the level of NUTS3 as input to the estimations. Finally in 2011, the experts of state statistics agreed that it is worth trying to integrate the estimation procedures into the system of official statistics. Folie 9

10 Methodological improvements of UA estimations from the Micro-Census At first, all estimations had to be based on the results for RAS. Only a few cities and LUZ were RAS themselves; therefore, RAS results had to be broken down by NUTS3 and put together again for the LUZ. Until ref.year 2003 (UA II), the breakdown was done in proportion to the population, the economically active or the number of dwellings. A first improvement for 2004 (UA III) was the observation of structural differences within the RAS leading to adjustments of the calculations. The essential step forward came with UA IV for ref. years : Annual NUTS3 results became available as input to the calculations. Estimations could be based on data for 5 years, thus reducing the sampling error. Analysing and smoothing the time series helped to validate the data. A typology of NUTS3 data was used to reveal structural differences and to derive corrective factors for territorial units with small sample sizes. Folie 10

11 Estimation of Labour market variables This estimation is based mainly on register data in particular from the Federal Employment Agency that provides the territorial distribution and the weights of the territorial units. Not included are civil servants and self-employed. Here, the Micro-Census provides the target values for the Federal States. These values are broken down to the NUTS3 and LAU2 level according to employed and unemployed persons registered by the Federal Employment Agency (BA). They are adjusted to the ILO-definitions by addition of civil servants (Beamte), from official statistics estimates of self-employed based on the Micro-Census. Folie 11

12 Calculation of data for NUTS 3 from MC Calculation of 4-year ( ) averages of final MC-results as basis for a cluster-analysis for each group of variables (such as households, dwellings, income etc.) for Germany, federal states, RAS, NUTS 3 Cluster analysis with 9 types of cluster for each variable (household, dwelling, education, income, commuting) for each NUTS3 with average value of each cluster Classification of each NUTS3 units according to its types of clusters Summation of the values of NUTS3 for each RAS 1. City equals RAS no further calculation is necessary 2. NUTS3 does not equal RAS: a. Smoothed and adjusted value for NUTS3 if more than 2500 original values per year b. Smoothed and adjusted value for RAS if less the 2500 original values per year Folie 12

13 Household Income the effect of smoothing and fitting H o u s e h o l d i n c o m e ( E U R, M e d i a n ) M C - o r i g i n a l H o u s e h o l d i n c o m e ( E U R, M e d i a n ) a f t e r s m o o t h i n g a n d f i t t i n g H e i l b r o n n, K S H e i l b r o n n, K S H e i l b r o n n, L K H e i l b r o n n, L K R A S R A S B a d e n - W ü r t t e m b. B a d e n - W ü r t t e m b e r g D e u t s c h l a n d D e u t s c h l a n d Folie 13

14 Smoothing Calculate EMA twice: forward and backward and then calculate the mean Folie 14

15 Territorial levels of Micro-Census data for the Urban Audit Territorial level provided as... applied for UA as... NUTS0: Germany as a whole: official results final NUTS1: Federal States: official results final NUTS2: State Gov. Districts: official results not used for UA MC Regional Adjustment Strata: ~ official results smoothed and fitted NUTS3: Districts (Kreise): Sub-strata / large LAU2 units: input to estimations potential input to estimations smoothed and fitted in future: smoothed and fitted Folie 15

16 Next steps It is the aim of KOSIS-Gemeinschaft Urban Audit to integrate the estimates from the Micro-Census and from federal labour market statistics into the system of official statistics. Experts from the Federal and from State Statistical Offices support the intended integration. Current tests should prove the reliability of the results for NUTS3 and promote the integration. The revision of social statistics in the ESS may also open perspectives of a better use of Micro-Census results for smaller territorial units. Folie 16

17 Households in Germany 1-person-households - household-net income - monthly rent per m2 for NUTS 3 in Germany Source: KOSIS-Gemeinschaft Urban Audit, Calculations from Labour force survey Folie 17

18 3. Further Developments Aim: General Frame for all household statistics in Germany and in Europe such as Micro-Census, Labour Force Survey, EU- SILC, ICT Timeframe: Starting in 2017 Folie 18

19 Household Surveys - Status quo Questions Households Households new selection Households LFS < year Polls LFS < year MC / LFS LFS polls less than a year EU-SILC ICT Permanent Sample (DSP) Total Folie 19

20 Basic model subsample of voluntary household surveys Core programme MC/LFS (especially short-term economic development) MZ additional programme Allocation to subsample: a priori Total of 220,000 households Voluntary response Modules EU-SILC IKT Other (MC, LFS, MC additional programme, EU-SILC, EU core variables) ICT EU-SILC Allocation to subsamples: two stage Allocation to modules Stratified/disproportionate random sampling two-stage Other surveys / 7 Households List of questions Folie 20

21 4. Conclusions Official statistics as a source for a comprehensive statistical picture of living and working conditions High quality of data Specific data are available at regional level Further regionalisation is needed Priority setting for further developments Resources for further regionalisation Folie 21

22 More information nales/kreiszahlen html =false&context=regatlas01 Folie 22

23 Thank you for Listening! Dr. Susanne Schnorr-Bäcker Klaus Trutzel Folie 23