The Dutch Virtual Census of 2001 A New Approach by Combining Different Sources Eric Schulte Nordholt ECE Census meetings Geneva, November 2004

2 Contents Introduction Census Data sources Combining data sources: micro-linkage Combining sources: micro-integration Social Statistical Database (SSD) Census tables History of the Dutch Census Comparison with Censuses in other countries Conclusions

3 Introduction Census Why a Census ? Statistical information for research and policy purposes What kind of information ? Size of (sub)population(s) Demographic and socio-economic characteristics, at national and regional level Gentlemen’s agreement Eurostat: co-ordinator of EU, accesion and EFTA countries in the 2001 Census Round Census Table Programme, every 10 years

4 Data sources Registers: Population Register (PR), 16 million records demographic variables: sex, age, household status etc. Jobs file, employees, 6.5 million records, and self-employed persons, 790 thousand records dates of job, branch of economic activity Fiscal administration (FIBASE) jobs, 7.2 million records, and pensions and life insurance benefits, 2.7 million records Social Security administrations, 2 million records, auxiliary information integration process Surveys: Survey on Employment and Earnings (SEE), 3 million records, working hours, place of work Labour Force Survey (LFS), 2 years: records education, occupation, (economic) activity

5 Combining sources: micro-linkage Linkage key: Registers Social security and Fiscal number (SoFi), unique Surveys Sex, date of birth, address (postal code and house number) Linkage key replaced by RIN-person Linkage strategy Optimizing number of matches Minimizing number of mismatches and missed matches

6 Combining sources: micro-integration Collecting data from several sources  more comprehensive and coherent information on aspects of person’s life Compare sources - coverage - conflicting information (reliability of sources) Integration rules - checks - adjustments - imputations Optimal use of information  quality improves Example: job period vs. benefit period

7 Social Statistical Database (SSD) Social Statistical Database (SSD): Set of integrated micro-data files with coherent and detailed demographic and socio-economic data on persons, households, jobs and benefits No remaining internal conflicting information SSD-set: Population Register (back bone) Integrated jobs file Integrated file of (social and other) benefits Surveys, e.g. LFS Combining element: RIN-person

8 Census tables (1) Preliminary work before tabulating Census Programme definitions: not always clear and unambiguous, e.g. economic activity Priority rules (characteristics of) main job (highest wage) employee or employer job or (partially) unemployed job or attending education job or retired engaged in family duties or retired age restrictions Tabulating register variables: simply straightforward counting from SSD-register data

9 Census tables (2) Tabulating survey (and register) variables Mass imputation? Pro’s: reproducible results Con’s: danger of oddities in estimates (e.g. high educated baby) Traditional Weighting? Pro’s: simple, reproducible results (if same micro-data and weights) Con’s: no overall numerical consistency between survey and register estimates Demand for overall numerical consistency 1 figure for 1 phenomenon all tables based on different sources (e.g. surveys) should be mutually consistent

10 Census tables (3), example Ethnicity: register Education: survey 1 and survey 2 Employment status: survey 2 Estimate: T1: educ x ethnic and T2: educ x employ Survey 1 Survey 2 Register ethnic 1...k educ Lo... Hi employ 1...m educ x ethnic not- NL NLTotal educ Lo educ Hi Total employ x educ employednon- employed Total educ Lo educ Hi Total RegisterSurvey 1 Survey Total NLnot- NL ethnic

11 Census tables (4) Repeated Weighting (RW) : tool to achieve numerical consistency (VRD-software) Basic principles of RW: estimate table on most reliable source (mostly source with most records, e.g. register) estimate tables by calibrating on common margins of the current table and tables already estimated (auxiliary information) repeatedly use of regression estimator: - initial weights (e.g. survey weights) calibrated as minimal as possible - lower variances - no excessive increase of (non-response) bias (as long as cell size>>0) each table own set of weights

12 Census tables (5), example continued Survey 1 Survey 2 Register ethnic 1...k educ Lo... Hi employ 1... m RegisterSurvey 1 Survey 2 educ x ethnic not- NL NLTotal educ Lo educ Hi Total employ x educ employednon- employed Total educ Lo educ Hi Total ethnicnot- NL NL Total ethnicnot- NL NL Total Calibrate on ethnic, then on educ x ethnic

13 History of the Dutch Census TRADITIONAL CENSUS Ministry of Home Affairs: 1829, 1839, 1849, 1859, 1869, 1879 and 1889 Statistics Netherlands: 1899, 1909, 1920, 1930, 1947, 1960 and 1971 Unwillingness (non-response) and reduction expenses  no more Traditional Censuses ALTERNATIVE: VIRTUAL CENSUS 1981 and 1991: Population Register and surveys development 90’s: more registers → 2001: integrated set of registers and surveys, SSD

14 Comparison with Censuses in other countries Traditional Census (complete or partial enumeration): Most countries (Estonia, Slovenia, Greece and the UK) Mixture of traditional Census and Registers: Some countries (Norway and Switzerland) Entirely or largely register-based Census: A few Nordic countries (Sweden and Finland) Virtual Census: The Netherlands Tables: /census-2001.htmhttp:// 2001/census-2001.htm Book: htmhttp:// htm

15 Conclusions The Dutch Virtual Census 2001 was successful with its innovative approach: new source: SSD, integration of registers and surveys (micro-integration remains important) new methodology for consistent estimation was implemented Pro’s: relatively cheap (cost per inhabitant) and quick Con’s: publication of small subpopulations sometimes difficult or even impossible because of limited information Solutions for Con’s : small area estimation (synthetic estimators)