Manuela Lenk Statistics Austria Registers Classifications and Methods

Manuela Lenk Statistics Austria Registers, Classifications and Methods Division 22 nd – 23 rd May 2012 www. statistik. at Quality assessment of register-based census data in Austria UNECE Expert Group Meeting on Censuses Using Registers, Geneva We provide information

Evolution of the Austrian census 2001 Last traditional census High burden for respondents, considerable financial effort (72 million Euros) Register-based Census 2011 -- The Movie - You. Tube 2006 Register-based test census Methods, data procedures and use of registers were successfully tested 2009 Start of annual register-based labor market statistics Possibility for application and improvement of methods 2011 First complete register-based census No burden for respondents, costs reduced to 10 million Euros – Data from 7 base and 7 comparison registers (35 data holders) www. statistik. at slide 2 | 23 May 2012

The principle of redundancy Ø Ø Same attribute in more than one register Comparison registers are used to confirm the values in the base registers due to quality improvement PIN SEX_CPR SEX_CSSR SEX_TR SEX_UR SEX_VALID RULE … … … 1 1 1 0 1 1 ID 3458 1 0 0 0 1 2 ID 3459 0 2 2 3 ID 3460 2 2 1 1 2 4 … … … … ID 3457 CPR=Central Population Register, CSSR=Central Social Security Register, TR=Tax Register, UR=Unemployment Register www. statistik. at slide 3 | 23 May 2012

Residence Analysis Persons who are only recorded in the CPR receive a letter containing the question: “Did you have your main residence in Austria at the reference date? – Yes or No? ” b. PIN OS CPR CSSR TR UR SWR CAR … … … … … … ID 3458 ID 3459 ID 3460 … … … ID 3457 CPR=Central Population Register, CSSR=Central Social Security Register, TR=Tax Register, UR=Unemployment Register, SWR=Register of Social Welfare Recipients, CAR=Child Allowance Register www. statistik. at slide 4 | 23 May 2012

Quality framework www. statistik. at slide 5 | 23 May 2012

Quality Framework – Documentation Ø Hyperdimension Documentation: HDD Focus on factors that possibly predetermine data quality. The reliability of the data source is checked. Ø Questionnaire for register authorities e. g. Is the attribute relevant for the data source keeper? Are technical input checks applied? How fast are changes edited in the register? ØQuality measure Each question is scored. The sum of scores in the questionnaire is compared with theoretical maximum score. Quality indicator = www. statistik. at obtained score maximal obtainable score slide 6 | 23 May 2012

Quality Framework – Pre-Processing Ø Hyperdimension Pre-Processing: HDP Focus on the data editing process. Definition and range errors, as well as missing primary keys and item non-response are detected. Total number of records - Records without unique key - Item non-response (with unique key) - Values out of range = Number of usable records ØQuality measure The quality measure gives the ratio between the number of usable records an the total number of records. Quality indicator = www. statistik. at usable records total number of records slide 7 | 23 May 2012

Quality Framework – External Source Ø Hyperdimension External Source: HDE Focus on the data accuracy and consistency. The register-based census data is compared with survey data (Microcensus). If the attribute of interest is not available in the Microcensus → Expert interview ØQuality measure The number of consistent values are compared with the total number of linked records. Quality indicator = www. statistik. at consistent values total number of linked records slide 8 | 23 May 2012

Register Level – Quality Measures Ø Overview: Quality measures of the three hyperdimensions Ø obtained score Documentation: hd. D = maximal obtainable score hd. P = usable records total number of records Ø Pre-processing: Ø External source: hd. E = total number of linked records Ø The overall quality indicator for each attribute in each register is a weighted sum of the three quality measures. consistent values v. D*hd. Dij www. statistik. at v. P*hd. Pij v. E*hd. Eij qij slide 9 | 23 May 2012

Quality measures for raw data from 2008 Register Attribute HDD HDP HDE q(33, 33) REG 1 SEX 1. 000 0. 998 0. 999 REG 2 SEX 0. 792 0. 942 0. 999 0. 911 REG 3 SEX 0. 444 0. 746 0. 997 0. 729 REG 4 SEX 0. 792 0. 993 1. 000 0. 928 REG 3 FT/PT 0. 381 0. 698 0. 847 0. 642 REG 5 EDU 0. 928 0. 950 0. 800 0. 891 www. statistik. at slide 10 | 23 May 2012

Quality framework www. statistik. at slide 11 | 23 May 2012

Census Database (CDB) - Attributes Ø Unique Attributes (C) Attribute exists in only one register, directly transferred to the CDB (e. g. highest level of education) Ø SEX_Reg 3 Multiple Attributes (A) Attribute exists in more than one register, combined in the CDB using certain decision rules (e. g. demographic attributes) Ø SEX_Reg 1 SEX_Reg 2 Derived Attributes (F and G) Attribute is created based on other attributes (e. g. family and household status) www. statistik. at Multiple Attribute Attrib 1 Derived Attribute Attrib 2 slide 12 | 23 May 2012

CDB – Quality assessment of multiple attributes Ø Average quality on register-level consistent & conflicting values are ignored Ø Application of Dempster-Shafer Theory Degree of belief is taken into account • Consistency with other sources: q. CDB ↑ • Conflict with other sources: q. CDB ↓ SEX: www. statistik. at q. REG 1, SEX = 0. 9 q. REG 2, SEX = 0. 7 PIN REG 1 REG 2 CDB AV q. CDB DS q. CDB 9845 male 0. 80 0. 99 4866 male female 0. 80 0. 77 2047 female 0. 80 0. 77 slide 13 | 23 May 2012

Potentials and perspectives Ø Quality indicators on register level - each attribute in each register is associated with a quality indicator - indicators can also be used for other projects and deliver essential information about the quality Ø Quality indicators for the Final Data Pool of register-based statistics - indicator for each attribute which covers all quality-related aspects quality of the register quality of imputations - The experience gained with the new census type can be of use for other population and household surveys - The quality framework can be used for other register-based statistics www. statistik. at slide 14 | 23 May 2012

Further Information Ø Austrian Journal of Statistics, Volume 39 (2010), Number 4 • Ø Statistica Neerlandica, Volume 66 (2012), Issue 1 • Ø http: //live. unece. org/fileadmin/DAM/stats/documents/ece/ces/ge. 41/2010/wp. 4. e. pdf European Conference on Quality in Official Statistics 2010, Helsinki • Ø http: //www. cros-portal. eu/sites/default/files/S 13 P 1. pdf UNECE/Eurostat Expert Group Meeting on Register-Based Censuses 2010, The Hague • Ø http: //isi 2011. congressplanner. eu/pdfs/650199. pdf NTTS Conference 2011, Brussels • Ø http: //www. ine. es/e/essnetdi_ws 2011/ppts/Lenk. pdf ISI World Statistics Congress STS 50 - Methods and quality of administrative data used in a census 2011, Dublin • Ø http: //onlinelibrary. wiley. com/doi/10. 1111/j. 1467 -9574. 2011. 00506. x/pdf ESSnet on Data Integration 2011, Madrid • Ø http: //www. stat. tugraz. at/AJS/ausg 104/104 Berka. pdf http: //q 2010. stat. fi/media//presentations/session-26/fiedler_quality-in-official-statistics_statisticsaustria_paper. pdf European Conference on Quality in Official Statistics, June 2012 (forthcoming) www. statistik. at slide 15 | 23 May 2012

Please address queries to: Manuela Lenk Register based census Contact information: Guglgasse 13, 1110 Vienna phone: +43 (1) 71128 -8283 fax: +43 (1) 71128 -7445 manuela. lenk@statistik. gv. at www. statistik. at Quality assessment of register-based census data in Austria UNECE Expert Group Meeting on Censuses Using Registers, Geneva slide 16 | 06 December 2020