Transcript Presentation - Quality on Statistics 2010

Using administrative
registers in sample surveys
European Conference on Quality in Official Statistics
3-–
6 May 2010
Kaja Sõstra
Statistics Estonia
Outline
Administrative data are used during different stages
of sample survey:
• establishing the survey frame
• sample design
• pre-filling of survey questionnaire
• calculating weights
• imputation
• small area estimation
Kaja Sõstra
Examples of sample surveys and
frames in Statistics Estonia
Statistical
unit
Person /
household
Survey
Frame
EU-SILC, LFS, HBS
Population register
Enterprise / Wages and salaries,
establishment SBS, Information
technology
Statistical business
register
Agricultural
holding
Structure of agricultural Statistical register of
holdings, Livestock
agricultural holdings
production
Kaja Sõstra
Statistical registers
Statistics Estonia has two main statistical registers
based mainly on administrative sources:
Statistical business register
• Commercial register
• Register of taxable persons
• other registers
Statistical register of agricultural holdings
• Register of agricultural support and land parcels
• Register of taxable persons
Kaja Sõstra
Sample design (1)
• Samples for surveys of economical units are selected
every year in November
• Administrative data is used for
– Stratification (economic activity, size class)
– Sample allocation (turnover, number of
employees)
– Detection of outliers
Kaja Sõstra
Sample selection (2)
• Samples for household surveys are selected in the
end of each year by Population Register
• Statistics Estonia gives conditions for
– establishing frame (age limits),
– stratification (sex, age, place of residence),
– sample size by stratas
• Sample is drawn using stratified systematic sampling
• Sample information is periodically updated by
Population Register (deaths, migration)
Kaja Sõstra
Problems with frames
• Lack of data for establishing correct frame
– Immigrant population includes persons who´s
parents were born abroad. Data about parents
was missing for about 80% of persons
– Data about institutions is missing
• Incorrect contact information in Population Register –
incorrect addresses cause high nonresponse and
increase cost of fieldwork
Kaja Sõstra
Pre-filling of questionnaires
• New development in Statistics Estonia
• Decreases response burden, saves time of
respondent
• Mainly the data from previous period of the same
survey is used
• New developments for using administrative data:
– annual bookkeeping reports for SBS from
Commercial register
– Data of agricultural support and land parcels
Kaja Sõstra
Calculation weights
Calculation weights consists of three steps
• Design weight
• Nonresponse adjustment weight
• Calibration weight
Administrative data is used during all steps
Kaja Sõstra
Nonresponse adjustment
• Nonresponse is increasing problem in household
surveys
• Usually only administrative data is known for
nonrespondents
• Response probability is estimated for every
household or person
• Design weight is adjusted according to estimated
response probability
Kaja Sõstra
Response rates by age and sex
100
%
80
60
40
20
0
15-64
65-74
Male
Average
75-84
15-64
65-74
75-84
Female
Harju and Rapla county
Kaja Sõstra
Population size and survey estimates by
age group, males
60
Thousand
50
40
30
20
10
Population size
80-84
75-79
70-74
65-69
60-64
55-59
50-54
45-49
40-44
35-39
30-34
25-29
20-24
15-19
0
Estimated number of persons
Kaja Sõstra
Calibration
• Calibration of weights is a final step of weighting in all
household surveys
• First trial to calibrate weights of agricultural survey
has done
• Calibration groups in labour force survey:
– Sex x 5-year age group (24 groups)
– Place of residence (15 counties and capital)
– Place of residence (urban, rural)
– Nationality (Estonian, non-Estonian)
Kaja Sõstra
Imputation
• Administrative data is used for cold-deck imputation
• Surveys of economic units:
– VAT, social insurance, annual bookkeeping
reports
• Household surveys
– personal income tax for imputation income
variables
Kaja Sõstra
Conclusions
• Good quality of administrative registers is important to
ensure quality of sample surveys
• There are plans to increase use of administrative data
for sample surveys in Statistics Estonia
– Establishing statistical population register
– Pre-filling of questionnaires
– Collecting part of data from registers
• Continuous development and updating administrative
registers and cooperation is important for improving
quality of all statistical activities
Kaja Sõstra
Thank you
for your attention!
Kaja Sõstra