Comparing approaches of different (partly) register-based

countries

Eric Schulte Nordholt

Senior researcher and project leader of the Census

Statistics Netherlands

Division Socio-economic and spatial statistics

e.schultenordholt@cbs.nl

Presentation at the Expert Group Meeting on Censuses Using Registers in Geneva (22-23 May 2012)

2

Contents

• Introduction to the research project

• Participating other countries

• Letter for participation• Census approaches of fully register-based countries• Census approaches of partly register-based countries• Conclusions

33

Introduction to the research project

Master’s thesis project of Martina Maris (University of Groningen) in the period September 2011 – April 2012 supervised by Mark Huisman (University of Groningen), Eric Schulte Nordholt (Statistics Netherlands) and Jantien van Zeijl (Statistics Netherlands) concentrating on a number of questions:

1. Which methods are used to compensate for missing information in the Census of 2011?

2. Which methods are used to combine register and survey data in the Census of 2011?

4

Participating other countries

Fully register-based countries• Norway• Denmark• Sweden• Finland• Austria• Slovenia

Partly register-based countries• Switzerland• Germany• Poland

4

5

Letter for participation

• Which variables are not available from registers?• What approach will be used when required variables

cannot be obtained from available registers?• Will sample surveys be used to obtain additional

census information?• Which estimation method will be used to combine

register and survey data?• What are the reasons for using this method?

• Large variation in answers

5

66

The Norwegian Census Approach

Approach used to obtain all required information:• Imputation (e.g., cohabitation)• some categories will not be specified (e.g.,

family workers)• persons will be assigned to a different category

(e.g., private households)

Approach used to combine register and survey data:• no sample surveys• quality assessment

7

The Danish Census Approach

Approach used to obtain all required census information:• additional survey data (e.g., educational

attainment)• some variables will be provided not completely

in accordance with their required definitions (e.g., CAS)

Approach used to combine register and survey data:• survey data on educational attainment for

immigrants • non-respondents imputed on the basis of

register data7

88

The Swedish Census Approach

Approach used to obtain all required census information:• imputed as existing (e.g., water supply)• provided not completely in accordance with the

required definitions (e.g., CAS)

Approach used to combine register and survey data:• no survey data• quality assessment

9

The Finnish Census Approach (1)

Approach used to obtain all required census information:• “register estimation method“

• estimate for each statistical unit the value of a target variable

• not all required categories will be specified (e.g., own-account worker of status of employment)

• quality problems (e.g., no information on students who have completed education abroad)

Approach used to combine register and survey data:• No survey data: all variables will be created by using the

register estimation method

9

1010

The Finnish Census Approach (2)

Classifying the main type of activity in 19 steps1. Persons aged 0-14 years below national

minimum age2. Persons aged 75 years and over

pensioners3. Unemployed

• persons aged 15-64 who, according to the Ministry of Employment and the Economy’s Register of Job Applicants are unemployed on the last working day of the year

[…]19. Others outside the labour force

• persons who do not meet any of the above criteria

1111

The Austrian and Slovenian Census Approaches

Approach used to obtain all required census information:• Imputation of all missing information

Approach used to combine register and survey data:• No survey data

1212

The Swiss Census Approach

Approach used to obtain all required census information:• An annual structural survey of 200,000 persons

(2.7%)• population, households, families, housing,

employment, mobility, education, language, and religion

• Thematic surveys of 10,000 to 40,000 persons • 1) mobility and transport, 2) education and

training, 3) health, 4) families and generations, and 5) language, religion and culture.

• An annual omnibus survey of 3,000 persons• provide quick answers to questions from policy

makers

Approach used to combine register and survey data: • Traditional weighting

1313

The German Census Approach

Approach used to obtain all required census information:• household survey (10% of population)

• education, employment, and migration background• traditional enumeration of buildings and housing (100%)

• year of construction, type of building, number of rooms

• traditional enumeration of all residential establishments and collective living quarters (100%)• boarding schools, homes for elderly people, • data necessary to identify the persons living in the

institutions (e.g., name, date of birth, place of birth, and sex)

Approach used to combine register and survey data:• Traditional weighting

1414

The Polish Census Approach

Approach used to obtain all required census information:• Survey for basic demographic information• Sample survey for detailed information

• 20% of the dwellings of permanent or temporary residents of the Republic of Poland

Approach used to combine register and survey data:• Traditional weighting

15

Conclusions

Several approaches are used to obtain all required census information• Most countries provide some of the data not completely

in accordance with required EU definitions (e.g. CAS)Imputation of missing information is a common approachMost used approach to combine register and survey data:• Traditional weighting

Repeated weighting is in the Dutch Census of 2011 only used for the variables occupation, level of education and possibly some housing variablesStatistics Netherlands calls its approach a virtual census: fully register-based in the sense that no census questionnaires exist, partly register-based in the sense that already existing survey information is used for a few important variablesThis study is of used for future recommendations on censuses