Title
2007 Adult Education Survey
Subtitle
Pilot Survey
"The first wave of the survey, 2007 AES (also called ‘pilot survey') was carried out between 2005 and 2008 in 29 countries. This pilot exercise was set up within a common EU framework including a standard questionnaire, tools and quality reporting. The reference year was set at 2007." (http://ec.europa.eu/eurostat/web/microdata/adult-education-survey)
The 2007 AES anonymised microdata includes 26 national datasets out of the 29 countries which implemented the survey. The total net sample size for the 26 countries is about 183 000 individuals.
Geographical Coverage
Austria, Belgium, Bulgaria, Croatia, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Italy, Latvia, Lithuania, Netherlands, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden, United Kingdom and Norway.
Ireland, Luxembourg, Liechtenstein and Iceland did not take part in the 2007 AES data collection. Although Malta, Switzerland and Turkey implemented the survey, the data from these three countries have been excluded from the anonymised dataset because the national authorities did not allow for this dissemination. (http://ec.europa.eu/eurostat/web/microdata/adult-education-survey)
Time Period Covered
2005-2008
-
Country Specific Information: AES 2007
-
LV - Latvia
Corresponding National Study
-
Producer
Central Statistical Bureau of Latvia
Universe
The target population are individuals aged 25-64, permanently resident in households.
Cross-sectional Data
Achieved Sample Size
2.314
- Cross-sectional
Source Of Sampling Frame
Central Statistical Bureau of Latvia (CSB) Address Register
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Area units
Secondary Sampling Units
Individuals
Stratification Criteria
stratum groups include 4 urbanisation levels (Capital city, 6 biggest cities, other cities and rural areas), 3 age groups (25-34; 35-49; 50-64)
Sampling Method
Within each stratum the selection of primary sampling units is made in the following steps:
• The primary sampling units are listed by geographical region, and within a geographical region in a serpentine order that places units containing similar types of people together.
• Cumulative totals of the size measure (the number of people estimated from census data) of primary sampling units are formed throughout each stratum.
The first stage PSU was selected by simple random sampling taking in to account stratification. On second stage using stratification in stratum groups individuals was selected using simple random sampling.
Final number of stratums were 2052. Number of selected individuals are 4611.
Start Date
05-2007
End Date
06-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
- 0 % Mixed Mode Collection
Percentage Of Proxy Interviews
0 %
Weighting: Method
The initial weights are calculated according to the sample design, with all persons within the same PSU having equal design weights. The design weights are adjusted using the data of response level in each PSU. In the next step adjusted weights was adjusted using calibration. For calibration was used auxiliary information. Calibration was calculated by 64 subgroups: 4 urbanisations levels; 2 sex groups and 8 age groups. Calibrated weights were calculated with the use of the specially created program in the software R.2.5.1.
Resources
-Eurostat: Adult Education Survey. Final Report Latvia.
-
LT - Lithuania
Corresponding National Study
-
Producer
Statistics Lithuania
Universe
The target population are individuals aged 25-64.
Cross-sectional Data
Achieved Sample Size
3.696
- Cross-sectional
Source Of Sampling Frame
Lithuania Population Register
Sampling Design
Probability
- random
- not stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Sampling Method
In the case the interviewer finds that individual selected does not live any more at the address indicated in the frame, this individual is changed by a member of a household living at the same address and belonging to the survey population by a rule of the nearest birthday. This last step is not taken into account when constructing the weights.
Start Date
03-2006
End Date
04-2006
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
- 0 % Mixed Mode Collection
Percentage Of Proxy Interviews
0 %
Coverage Adjustment Weight: Method
The groups of individuals used for calibration are: people
1) living in the county (10 counties, consequently, 10 groups);
2) belonging to the age groups: 25-34, 35 - 49. 50-54 years old;
3) being a man;
4) living in the rural region;
5) being a women and living in the urban region.
Weighting: Method
SAS macro program CLAN of Statistics Sweden has been used for calculation of the calibrated weights and estimates of the variances of the estimators.
Resources
-Eurostat: Adult Education Survey. Final Report Lithuania.
-
HR - Croatia
Corresponding National Study
-
Producer
Central Bureau of Statistics
Universe
The target population are individuals aged 25-64.
Cross-sectional Data
Achieved Sample Size
3.089
- Cross-sectional
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Households
Secondary Sampling Units
Individuals
Stratification Criteria
The sample is implicitly stratified according to the size of settlements and according to the region (21 counties). At the first stage 500 sampling units are selected with the probability proportional to size (PPS with replacement), and at the second stage using random procedure 10 persons aged between 25 and 64 years are selected. All of them were from different households. Thus 5000 persons were selected.
Sampling Method
In the sample selection process full name and address of citizens were used as well as information about gender, date of birth and education level. Respondents (persons aged between 25 and 64 years) were randomly selected from this data base. For the allocation of respondents, in addition to the Data Base of individual Census data, also the frame of the Census Enumeration Units is used. These units serve as a sampling frame for the selection of units at the first stage in two-stage sampling design.
Start Date
12-2007
End Date
12-2007
Participation Mandatory
Yes
Type Of Data Source
Survey
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
Rim-weighting procedure (iterative proportional fitting) was used for grossing-up individuals. Variables on which grossing-up procedure was based on were as follows: region, settlement size, gender, age and level of education, i.e. extrapolation weight was calculated for each participant in the survey and calculation method has taken into account participants region, settlement size, gender, age and level of education. This procedure automatically takes into account sampling method and unit non-response.
Resources
-Eurostat: Adult Education Survey. Final Report Croatia.
-
RO - Romania
Corresponding National Study
-
Producer
National Institute of Statistics Romania
Universe
The target population are individuals aged 25-64 in private households.
Cross-sectional Data
Achieved Sample Size
13.909
- Cross-sectional
Source Of Sampling Frame
master sample EMZOT: a Multifunctional Sample of Territorial Areas, made by the data registered in the Population and Dwelling Census (in 2002)
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Area units
Secondary Sampling Units
Dwellings
Stratification Criteria
County (41) and urban/rural area. The six sectors of Bucharest correspond to separate strata. In total 88 strata.
Sampling Method
In a first phase, the national territory was delimitated in geographical areas, named Primary Units (PU), so that no area of the national territory was excluded. From these PU was extracted a sample of PU which constitutes the research centers in EMZOT. From these research centers are selected the dwellings samples for household surveys.
A primary unit (PU) was constituted of minimum three neighboring census sections, in order to fulfill the condition of a minimum necessary volume calculated as number of permanent dwellings (1400).
As the accuracy of the estimations resulted from the surveys depends, mostly, on the quality of PU constitution and selection, in view of ensuring an optimal representativeness at national and regional level, the definition of UP was made on layers, constituted on the basis of the criteria county and area of residence (urban, rural).
Thus, it resulted a number of 88 layers (in the Municipality of Bucharest, the 6 administrative districts constitute distinct layers).
The probabilities of PU inclusion were calculated proportionally with PU size, which is defined as number of permanent dwellings (as only households from the permanent dwellings, in the moment of the study, are investigated, the other dwellings being considered outside the survey’s scope).
The extraction of the Primary Units in the first phase was made separately on each layer using the method of balanced extraction using an automatic procedure performed under SAS software (macro CUBE).
By using this method, the quality of PU extraction was considerably improved, thus influencing the quality of the survey results in the social statistics domain.
The management system of the EMZOT sampling frame is designed so that it enables the strict monitoring of the selections performed in time and space and the identification of certain characteristics required for the treatment of non-responses.In the second phase of sampling are selected, from each Primary Unit (survey centre), Secondary Units of sampling, represented by dwellings, based on the method of systematic selection. This method ensures a very good dissemination of the surveyed dwellings within the centre.
Start Date
03-2008
End Date
04-2008
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
14.62 %
Weighting: Method
Due to non-response in the second stage the design weights has to be adjusted for nonresponse.
The second stage design weights are adjusted in two steps.Step 1 – expansion with inverted response rates:
The second stage design weights were adjusted by expanding them with the inversed response rate in each survey centre.
Under the assumption that the households respond independent and with equal response probabilities within survey centres, estimator is an unbiased estimator of the total population.Step 2 – use of auxiliary information:
In order to approve the accuracy of the estimates, auxiliary information were used in a regression estimator.
In total 96 auxiliary groups are used in the regression estimator. The structure of the population variables mentioned is known from external sources.
Resources
-Eurostat: Adult Education Survey. Final Report Romania.
-
NO - Norway
Corresponding National Study
-
Producer
Statistics Norway
Universe
The target population are individuals aged 25-64, permanently living in Norway.
Cross-sectional Data
Achieved Sample Size
3.018
- Cross-sectional
Source Of Sampling Frame
Statistics Norway population database of the Central Population Register (DSF)
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Area units
Secondary Sampling Units
Individuals
Stratification Criteria
"geographical strata (stratified based on industrial structure, number of inhabitants, centrality, communication structures, commuting patterns, trade areas and (local) media coverage)
stratification by three age groups (25-34, 35-49, 50-64)
Sampling Method
In the first sampling stage 107 Primary sampling units were selected from geographical strata.
In the second stage individuals were sampled from these primary sampling units. The total number of sampled individuals was 4500. Statistics Norway’s standard two-stage probability sample design for face-to-face interviews is intended to assure equal selection probability for all sampled individuals. Due to the stratification by age this is not quite the case for the AES (the sampling plan does not adjust for the fact that the age distribution varies somewhat between the primary sampling units). Some adjustments were made in the estimation of final weights to reduce possible bias.
Start Date
05-2007
End Date
08-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 27 % CAPI
- 73 % CATI
Percentage Of Proxy Interviews
0 %
Weighting: Method
The estimator used in the Norwegian AES is a regression estimator, where auxiliary information has been used to reduce bias from sampling errors and non-response errors.
The population U is divided into non-overlapping groups, Up. The division is made through the help of auxiliary information collected from different registers. The variables used for the division is:
• Sex*age*level of education (ISCED 5+6, ISCED 3A+B+3C long+4, no or below ISCED 3C long) (30 groups)
• Region (7 groups)
Resources
-Eurostat: Adult Education Survey. Final Report Norway.
-
HU - Hungary
Corresponding National Study
-
Producer
Hungarian Central Statistical Office
Universe
The target population are individuals aged 25-64, living in private households in Hungary.
Cross-sectional Data
Achieved Sample Size
7.494
- Cross-sectional
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Households
Secondary Sampling Units
Individuals
Stratification Criteria
Stratum 1 consisted of individuals who had participated in adult education within four weeks prior to the LFS interview, and stratum 2 consisted of the rest of individuals aged 25-64.
The sample design strata were defined in terms of geographic units, size categories of settlements and area types such as city centres, outskirts, etc.
Sampling Method
With the aim of having a sample such that individuals involved in adult education are overrepresented, segments of the LFS sample from earlier quarters were used to define the sampling frame, since the LFS collects information on adult education. In particular, groups of individuals aged 25-64 and reporting in the LFS for the first time in quarters 2, 3 and 4 of 2005 and in quarter 1 of 2006 were identified. Taking the union of these groups as frame, two strata were defined. Finally, stratum 1 was completely included in the AES sample, while a simple random sample of 7700 individuals was selected from stratum 2.
Start Date
07-2006
End Date
09-2006
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
Sample weights for the AES were defined by the following rule: design_weight_of_the_AES = sampling_fraction x final_weight_of_the_LFS where sampling_fraction equals 1 or 7700/40949 = 0.188 in strata 1 and 2 , respectively.
The final sample weights of the LFS, which are the starting points of the design weights of the AES, are obtained as the result of a calibration process. Among other things they are adjusted so that estimated totals in 2x10 age-sex groups in 20 administrative units (counties) may agree with the corresponding population values. The latter are updated census counts computed in each calendar month by a version of the demographic components method. The final weights of the AES are determined in a similar way, though – owing to smaller sample size – the total number of controls amounts only to 71. This means updated counts for 2x5 age-sex groups in 7 NUTS-II level regions plus an estimate of the number of individuals in stratum 1 in the population. The raking algorithm used for calibration in a SAS environment has produced rapid convergence and fairly acceptable (i.e. little) change in the value of the design weights.
Resources
-Eurostat: Adult Education Survey. Final Report Hungary.
-
BG - Bulgaria
Corresponding National Study
-
Producer
National Statistical Institute - Bulgaria
Universe
The target population are all individuals aged 25-64, permanently resident in Bulgaria.
Cross-sectional Data
Achieved Sample Size
2.825
- Cross-sectional
Source Of Sampling Frame
Population Census 2001; Register
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Households
Secondary Sampling Units
Individuals
Stratification Criteria
"The sample is “stratified” by administrative-territorial districts in the country (28 districts) separately for urban and rural population. As a result 56 strata were formed (28 strata of urban and 28 strata of rural population). In the frame of each stratum the municipalities are ranged according to the number of their population.
The number of clusters is calculated for stratum included in the sample.
Clusters on the first stage for each one of the 56 strata are chosen with probability proportional to the size of district’s population in the general universe. The total number of selected clusters is 798."
Sampling Method
When forming the sample a two-stage sampling on a territorial principle is designed as follows: at the first stage the census enumeration areas (clusters) are selected; at the second stage the households to be surveyed are identified.
Start Date
11-2007
End Date
12-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
To deal with non-response post-stratification was used. The weight of each response (interviewed individual) individual is calculated by dividing the population in each post-stratum by the number of the interviewed persons in the same post-stratum.
Resources
-Eurostat: Adult Education Survey. Final Report Bulgaria.
-
FR - France
Corresponding National Study
-
Producer
Institut National de la statistique et des Etudes Economiques
Universe
The target population are individuals aged 25-64, living in households, and part of the population living in collective dwellings.
Cross-sectional Data
Achieved Sample Size
15.350
- Cross-sectional
Sampling Design
Probability
- random
- not stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Sampling Method
One person was interviewed per household.
If more than one person from a household fell within the field of the survey, one person was randomly selected from the potential interviewees.
Start Date
01-2006
End Date
01-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
A correction for non-response is made for each quarter independantly, using several criteria: gender, age, employment status.
Resources
-Eurostat: Adult Education Survey. Final Report France.
-
BE - Belgium
Corresponding National Study
-
Producer
Statistics Belgium
Universe
The target population are individuals aged 25-64 living in private households.
Cross-sectional Data
Achieved Sample Size
4.850
- Cross-sectional
Source Of Sampling Frame
National Register
Sampling Design
Probability
- random
- stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
The 18 strata are defined by:
3 regions: Brussels, Flanders, Wallonia
3 age classes: 25-34, 35-49, 50-64
2 sexes
Sampling Method
No specific restrictions were imposed for individuals living in the same household (i.e. it is possible that more than one individual is selected from the same household ). Individuals living in institutions, hospitals etc. are excluded.
Start Date
02-2007
End Date
06-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 69 % Self-administered
- 31 % Websurvey
Percentage Of Proxy Interviews
0 %
Weighting: Method
Correction for unit non-response and calibration are made in one step by raking ratio. The adjustment is made on the following groups:
- 18 sampling strata : 3 regions x 3 age groups x 2 sexes
- 3 employment status (employed, not-employed, inactive)
- 3 ISCED categories [low (ISCED 0/1/2), medium (ISCED3/4), high (ISCED5/6)]
- 2 nationality categories: Belgians/foreigners
- 3 categories of household size: 1 person in household/2-6 persons/more than 6 persons
Nationality is an important explanatory variable for non response.
Population totals for employment status and ISCED categories are estimated from labour force surveys results 2008. The other totals are extracted from national population register.
Resources
-Eurostat: Adult Education Survey. Final Report Belgium.
-
DE - Germany
Corresponding National Study
-
Producer
TNS Infratest Sozialforschung GmbH
Universe
Target Population of German AES for Eurostat: Persons in private households, 25 to 64 years old. (Data for Germany cover additionally the age group 18 to 24 years)
Cross-sectional Data
Achieved Sample Size
6213
- Cross-sectional
Sampling Design
Probability
- random
- stratified
- multi-stage
Sampling Method
Three-stage sampling:
First stage: Selection of PSUs
Second stage: Selection of households using a random walk
Third stage: Random selection of an individual household member as respondent (FSU)
Start Date
01-03-2012
End Date
25-06-2012
Participation Mandatory
Yes
Type Of Data Source
Survey
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Resources
- -
DK - Denmark
Corresponding National Study
-
Producer
Statistics Denmark
Universe
The target population are individuals aged 25-64.
Cross-sectional Data
Achieved Sample Size
3.100
- Cross-sectional
Source Of Sampling Frame
Central Person Register (CPR) at the beginning of 2008
Sampling Design
Probability
- random
- stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
The population was divided into three non-overlapping strata, and samples were selected from each stratum independently. The three strata are defined by the variable age.
Stratum 1: 24 < age < 35
Stratum 2: 34 < age < 50
Stratum 3: 49 < age < 65
Sampling Method
The persons selected were checked for telephone number, and contacted by phone if possible. The persons where a phone number could not be identified were contacted by letter asking either to get in touch with Statistics Denmark by phone or alternatively to answer the interview on the web.
Start Date
05-2008
End Date
07-2008
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 70 % CATI
- 30 % Websurvey
Percentage Of Proxy Interviews
0 %
Coverage Adjustment Weight: Method
An important characteristic of the regression estimator is the calibration property. Since information from administrative registers is used as auxiliary information the final estimates are based on distributions from administrative registers and not from the response population. This means that the final estimates are calibrated according to the auxiliary information from administrative registers.
Weighting: Method
The weighting procedure in the AES survey is based on the regression estimator. In the regression estimator auxiliary information is exploited in the estimation procedure. In the case of the Danish AES survey the auxiliary information comes from administrative registers.
The basic idea in the regression estimator is that relevant auxiliary information (chosen in a way that it covariates with variables of interest in the survey) contain information about study variables. This information is utilised in the estimation process to obtain better estimates. The stronger the correlation between target variables and auxiliary variables, the better is the final estimate. The quality is improved by lowering the standard error on estimates and/or by correcting for non response bias.
The estimation is carried out using the SAS-macros CLAN developed by Statistics Sweden. The variance (and standard error) estimate is based on a Taylor approximation.
The variables used in the estimation are
Gender (2 categories)
Age (4 categories)
Education (3 categories)
Income (4 categories)
Socioeconomic status (6 categories)
Family type (4 categories)
Occupation (4 categories)
Resources
-Eurostat: Adult Education Survey. Final Report Denmark.
-
FI - Finland
Corresponding National Study
-
Producer
Statistics Finland
Universe
The target population are resident household population, aged 25-64.
Cross-sectional Data
Achieved Sample Size
4.162
- Cross-sectional
Source Of Sampling Frame
Population database 2006 based on the Population Information System by the Population Register Centre
Sampling Design
Probability
- random
- stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
25-34 n=.951
35-64 n=4.491
Start Date
03-2006
End Date
08-2006
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Design Weight: Method
Inverse of selection probability
Coverage Adjustment Weight: Method
In groups 25-34 and 35-64 the calibration classes were "education level&agegroup" (3x4) and "gender®ion" (2x18). In group 18-24 only post stratification "gender&greater region" (2x3) was conducted.
Weighting: Method
The results of the EU-AES are going to be weighted against the total population by means of sample weights. The weights are established by using the probability of each observation to be included in the sample.
The weights are computed by Calmar software, so that the estimated marginal distributions of certain background variables (region, person's age group and gender, ISCED) correspond to the marginal distributions in the whole population, i.e. the population structure. The weights are formed as a ratio between the estimated frequencies in population cells and the known sample frequencies. Small or even zero cell frequencies are disregarded in this procedure. Thus, the method can be considered an "incomplete post-stratification". The weights are obtained more or less automatically by SAS programs used in processing the results.
Resources
-Eurostat: Adult Education Survey. Final Report Finland.
-
AT - Austria
Corresponding National Study
-
Producer
Statistics Austria
Universe
The target population are individuals of the age-group 25 to 64 years, permanently resident in Austria.
Cross-sectional Data
Achieved Sample Size
4.675
- Cross-sectional
Sampling Design
Probability
- random
- stratified
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
The strata are defined by 3 age groups (25-34 years, 35-49 years and 50-64 years) and sex.
Sampling Method
A primary and a second sample was drawn both of which were comprised of 8.500. Each person in this second sample was assigned to a certain person of the first sample and therefore could serve as a replacement for neutral losses and refusals in the first sample. If the interviewers were not able to conduct an interview at the first address, they were allowed to make use of the connected address in the second sample. Thus the unit response is a combination of 78 % cases from the first sample and 22 % cases from the second sample.
Start Date
04-2007
End Date
11-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
0 %
Non-response Weight: Method
Hierarchical hot deck imputation was used for the variables ISCED2D and MAINSTAT where missing values were not allowed.
Weighting: Method
Weighting was performed in three steps. First in the 6 strata (3 age groups X sex: altogether 6 strata) the design weight was assigned according to marginal distributions of the Austrian Population forecasts, which serve also as key figures for the Austrian Labour Force Survey. In a second step Non response factors – also within the 6 basic strata were used to inflate the design weights. The final step was done by calibrating the basic weights (originating from the two previous steps).
This final calibration was done according to 4 dimensions:
1. Nuts (Federal provinces) X sex
2. 5 year age-groups
3. Educational level
4. Labour force status
Resources
-Eurostat: Adult Education Survey. Final Report Austria.
-
CZ - Czech Republic
Corresponding National Study
-
Producer
Czech Statistical Office
Universe
The target population are individuals of the age of 25-64 years, permanently resident in the Czech Republic.
Cross-sectional Data
Achieved Sample Size
9.543
- Cross-sectional
Source Of Sampling Frame
Czech census enumeration unit register
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Census enumeration units
Secondary Sampling Units
Households
Stratification Criteria
Region (NUTS3) and municipality size was used for stratification. Final number of stratums was 56.
Start Date
01-2008
End Date
03-2008
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
0 %
Non-response Weight: Method
Due to the limited information on non-respondents restricted only to the geographical information obtainable from the sampling frame, the possibilities for modelling using propensity to response models were quite limited. So no such methods were executed.
Weighting: Method
The calculations were implemented using the CALMAR software in SAS. We use linear truncated method with lower limit 200 L and upper limit 1900 U. Linear truncated method minimize function of distance G.
Resources
-Eurostat: Adult Education Survey. Final Report Czech Republic.
-
CY - Cyprus
Corresponding National Study
-
Producer
Statistical Service of Cyprus
Universe
The target population are individuals aged 25-64, permanently resident in Cyprus.
Cross-sectional Data
Achieved Sample Size
4.810
- Cross-sectional
Sampling Design
Probability
- random
- stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
The variables used for stratification were: District and Urban/Rural area. There are 4 districts broken down by urban/rural area and 1 district which is only rural, thus giving 9 strata (Lefkosia urban, Lefkosia rural, Ammochostos rural, Larnaca urban, Larnaca rural, Lemessos urban, Lemessos rural, Pafos urban, Pafos rural).
Sampling Method
One individual was selected from each household, randomly using the birth date as a criterion (the person aged 25-64 having his/her birth date closest to the interview date was selected).
Start Date
09-2006
End Date
12-2006
Participation Mandatory
No
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Non-response Weight: Method
The non-response rate was used as a correction factor in the weighting procedure. (The inverse of the non-response rate is multiplied with the design weight, thus giving us a corrected design weight.)
Coverage Adjustment Weight: Method
The calibration variables used were the distribution of the population by age ( 25≤age≤29, 30≤age≤34, 35≤age≤39, ..., 60≤age≤64) and gender.
Weighting: Method
The design weight of the household (number of households in the frame divided by the number of households in the sample) was corrected for Non-response by multiplying it with the inverse of the Non-response rate.
The Final Weight of the household was calculated by multiplying the corrected Design Weight with the number of persons aged 25-64 within each selected household.
Resources
-Eurostat: Adult Education Survey. Final Report Cyprus.
-
SE - Sweden
Corresponding National Study
-
Producer
Statistics Sweden
Universe
The target population are individuals aged 25-64, permanently living in Sweden.
Cross-sectional Data
Achieved Sample Size
3.632
- Cross-sectional
Source Of Sampling Frame
the Swedish total population register (TPR)
Sampling Design
Probability
- random
- stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
Sex, 3 age groups and level of education was used for stratification. Level of education was coded into three groups according to Swedish classification of education (SUN2000). Level of education was collected from the Swedish register over participation in formal education.
Sampling Method
The final number of first stage stratums was 18. The sample was allocated to fulfil precision requirements for pre-defined domains.
Start Date
10-2005
End Date
03-2006
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 63 % CAPI
- 37 % CATI
Percentage Of Proxy Interviews
0 %
Weighting: Method
The design weight, di, is calculated by: di=Nh/nh, where Nh is the population size in stratum h and nh is the sample size in stratum h.
The non-response weight, gi, is calculated from auxiliary information:
The population U is divided into non-overlapping groups, Up. The division is made through the help of auxiliary information collected from different registers. The variables used for the division is:
• Sex*age*level of education (30 groups)
• Country of birth (3 groups)
• Employment status (3 groups)
• Degree of urbanisation (3 groups)
Resources
-Eurostat: Adult Education Survey. Final Report Sweden.
-
SI - Slovenia
Corresponding National Study
-
Producer
Statistical Office of the Republic of Slovenia
Universe
The target population are individuals aged 25-64, living in private households.
Cross-sectional Data
Achieved Sample Size
4.192
- Cross-sectional
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Area units
Secondary Sampling Units
Individuals
Stratification Criteria
Stratification according to type of settlements (6 strata):
1. settlements with fewer than 2.000 inhabitants and with less than 30 % of agricultural households;
2. settlements with fewer than 2.000 inhabitants and with at least 30 % agricultural households;
3. settlements with 2.000 to 10.000 inhabitants;
4. settlements with 10.000 to 80.000 inhabitants;
5. Maribor (the second largest city in Slovenia with approx. 93.000 inhabitants);
6. Ljubljana (Slovenia’s capital with approx. 250.000 inhabitants).
Sampling Method
Within each stratum 900 sampling units were selected with the probability proportional to size (PPS with replacement) and then in each sampling unit 8 persons were selected. Thus 7200 persons were selected. Sample size is based on the regulation size of sampling errors and on the knowledge of response rates in recent social surveys.
Start Date
09-2007
End Date
09-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 36 % CAPI
- 64 % CATI
Percentage Of Proxy Interviews
0 %
Non-response Weight: Method
Non-response is different in different statistical regions and is also larger in urban than in rural settlements. Weights due to non-response were calculated according to strata. The weight due to non-response is inversely proportional to the response rate within the stratum. Correction due to non-response is equal to the product of the sampling weight and the non-response weight.
Coverage Adjustment Weight: Method
Calibration of weights was aided by external sources. The objective was to achieve accurate estimates, which means that weighted estimates should equal known population values. Registers as well as aggregates for certain variables in the database are available. Therefore the weights are calibrated to these values. For the calibration of weights the SAS Macro Calmar was used. For the calibration, the sampling weight due to unequal probabilities of selection and non-response was used. Weights were calibrated on the basis of the following:
1. distribution of the population in Slovenia on 30th June 2007 by 12 regions
2. sex and age distribution of the population in Slovenia on 30th June 2007 in predefined classes (age classes: 25-29, 30-34, 35-39, 40-44, etc.),
3. education by 3 ISCED levels from the Labour Force Survey 2007
4. employment status from the Labour Force Survey 2007
Weighting: Method
n persons among N persons in the sampling frame were selected and therefore the probability of selection would be n/N, or the sample weight for the selected person would be N/n. As oversampling of certain strata needed to be considered, the sampling weight for the person belonging to stratum h, h=1, 2,…,6, is calculated in the following way:
Nh/nh * correction due to oversampling in stratum hThe sum of final weights is equal to the size of the population of persons aged 25 to 64 years in Slovenia. On 30th June 2007 there were 1 162 228 inhabitants aged from 25 to 64 years.
Resources
-Eurostat: Adult Education Survey. Final Report Slovenia.
-
SK - Slovakia
Corresponding National Study
-
Producer
Statistics Slovakia
Universe
The target population are individuals aged 25-64, permanently living in Slovakia.
Cross-sectional Data
Achieved Sample Size
-
- Cross-sectional
Source Of Sampling Frame
Slovak total population, state to 31st December 2005
Sampling Design
Probability
- random
- not stratified
- multi-stage
Primary Sampling Units
Municipality
Secondary Sampling Units
Households
Stratification Criteria
By random simple sampling was selected the municipalities where the survey was made up. Within the frame of these municipalities were selected the households by random simple sampling and within the frame of these household were selected the individuals that were interviewed.
Start Date
08-2007
End Date
12-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
0 %
Resources
-Eurostat: Adult Education Survey. Final Report Slovakia.
-
UK - Great Britain
Corresponding National Study
-
Producer
National Centre for Social Research (NatCen) on behalf of the Department for Innovation Universities and Skills (DIUS).
Universe
The target population are individuals aged 25-64., living in private households in England, Wales and Scotland.
Cross-sectional Data
Achieved Sample Size
3.528
- Cross-sectional
Source Of Sampling Frame
Postcode Address File
Sampling Design
Probability
- stratified
- multi-stage
Primary Sampling Units
Postecode sectors
Secondary Sampling Units
Addresses
Stratification Criteria
Stratification by region, age, the social economic group of the head of household and the proportion of male unemployment according to the 2001 Census of the population. In England and Wales the region variable was Government Office Region. In Scotland the region variable was counties.
Sampling Method
• The PSUs were postcode sectors (or groups of sectors). Prior to selection the PSUs were stratified by region, age, the social economic group of the head of household and the proportion of male unemployment according to the 2001 Census of the population. In England and Wales the region variable was Government Office Region. In Scotland the region variable was counties.
• Specifically, the list of postal sectors was sorted into the 10 standard regions or 14 counties. Within each region/counties, sectors were listed in order of percentage of households with residents above pension age. Cut-off points were then drawn approximately one third and two thirds (in terms of DPs) down the ordered list, to create three roughly equal-sized bands. This was done separately for each region/county. Five counties were not split in this way because they were relatively small. Within each of the 30/32 bands thus created (10 x 3) or (9 x 3) + 5, sectors were listed in order of percentage of households in non-manual SEG (SEG groups 1-6 and 13). Cut-off points were then drawn approximately one third and two thirds (in terms of DPs) down the ordered list, to create three roughly equal-sized bands. This was done separately for each of the 30 initial bands in England and Wales, and for the largest 24 bands in Scotland. Within each of the 90/80 bands thus created (30 x 3) or (24 x 3) + 8, sectors were listed in order of percentage of males of working age who were unemployed.
• A total of 265 PSUs were then selected with probability proportionate to the number of addresses within them. 215 of these PSU were in England and Wales, 50 PSUs were in Scotland. The intention was to boost the number of Scottish PSUs to enable separate analysis to be carried out in Scotland.
• Within each PSU a random sample of 42 addresses were systematically selected. This gave a total of 11,130 issued addresses.
• At each issued address the number of households (or dwelling units) was enumerated by the interviewer. At addresses with more than one household a single household was selected at random by the interviewer. The selected household was then invited to participate in the interview using a Kish grid.
• Within each responding household the interviewer enumerated the number of eligible persons and selected one at random using a Kish grid. This person was then asked to complete the questionnaire.
Start Date
10-2005
End Date
02-2006
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
The first step of the weighting procedure was to generate a set of selection weights. These selection weights were the product of a household selection weight and an individual selection weight. The household weight corrects for unequal selection probabilities of households at addresses where there was more than one household present. In these instances one household was selected at random. The household weight is equivalent to the total number of households at the address. The individual selection weight corrects for the selection of individuals in households where there was more than one eligible adult. The individual selection weight is equivalent to the number of eligible individuals in the household.
The age and sex distribution within each Government Office Region (GOR) of the survey data (weighted for different selection probabilities using the weight described above) was adjusted to match the ONS mid-2004 estimates. This allows variation by GOR to correct for any disproportionate non-response across the GORs.
Resources
-Eurostat: Adult Education Survey. Final Report UK.
-
IT - Italy
Corresponding National Study
-
Producer
Istat - National Institute of Statistics
Universe
The target population are households, individuals of 3 years and over, individuals of 14 years and over , individuals of 18 years and over and occupied people.
Cross-sectional Data
Achieved Sample Size
27.848
- Cross-sectional
Source Of Sampling Frame
Public register of households
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Municipality
Secondary Sampling Units
Households
Stratification Criteria
During the survey in question, the municipalities were stratified according to their demographic size and observing the following conditions:
• self-deliberation of the sample on a regional level
• selection of a sample municipality within each layer defined on the municipalities of the NAR group
• choice of a minimum number of households to be interviewed in each sample municipality; for this survey, this number was set at 24
Sampling Method
Municipalities are divided into two sub-groups on the basis of the resident population:
• The group of Self-Representative municipalities (which since now we‘ll referred as AR municipalities) made up of larger demographic size of municipalities;
• The group of Non Self Representative municipalities (or NAR) made up of the remaining municipalities.
Within the group of AR municipalities, each municipality is considered as an independent layer and a pattern known as cluster sampling is used. The primary sampling units are represented by households registered in the Public Registry, taken systematically from the Public Register of the municipality itself; for each registered household included in the sample, the characteristics that are the subject of the survey for all the actual components that belong to the same household are extracted.
Within the group of NAR municipalities a two stage pattern with layering of the primary units is adopted. The Primary Units (UP) are the municipalities, the Secondary Units are the households registered in the Public Registry; for each registered household included in the sample, the characteristics that are the subject of the survey for all the actual components that belong to the same household are extracted
The municipalities are selected with probabilities that are proportional to their demographic size and without reintroduction, and the households are extracted with probabilities that are equal and without reintroduction.
Start Date
05-2006
End Date
05-2006
Participation Mandatory
Yes
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
10 %
Weighting: Method
post-stratification method according to sex, age and region
Resources
-Eurostat: Adult Education Survey. Final Report Italy.
-
PT - Portugal
Corresponding National Study
-
Producer
Statistics Portugal
Universe
The target population are individuals aged 25-64, living in dwellings of main residence in national territory.
Cross-sectional Data
Achieved Sample Size
9.854
- Cross-sectional
Source Of Sampling Frame
Master sample based on 2001 Census
Sampling Design
Probability
- random
- not stratified
- multi-stage
Primary Sampling Units
Area units
Secondary Sampling Units
Dwellings
Start Date
09-2007
End Date
12-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % CAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
For the estimation procedure we had into consideration the application of one extrapolator which was calculated in three steps:
1st step: determination of an initial extrapolator “a priori”, based on the Horvitz-Thompson estimator, in which the weights of each unit are given by their inverse probability of selection;
2nd step: application of a correction factor for non-answers (as the total non-response is taking part of the final extrapolator calculation) in the extrapolator calculated in the first step. This procedure intend to balance the effect of non answers in the sample size, which was the basis for the initial extrapolator;
3rd step: correction of the initial extrapolator applying the boundary adjustment method, in order of having a distribution of the effectives extrapolated by the values oh the variables considered in the adjustment similar to the structure to the correspondent universe.
Resources
-Eurostat: Adult Education Survey. Final Report Portugal.
-
PL - Poland
Corresponding National Study
-
Producer
Central Statistical Office, Poland
Universe
The target population are individuals aged 25-64.
Cross-sectional Data
Achieved Sample Size
13.225
- Cross-sectional
Source Of Sampling Frame
Domestic Territorial Division Register
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Census enumeration units
Secondary Sampling Units
Dwellings
Stratification Criteria
Stratification was carried out separately in each voivodship (NUTS2) with consideration to its specifics. Primary sampling units were stratified according to the size of a locality. In urban areas, census enumeration areas were aggregated according to the size of towns. Cities comprised independent strata, while in the five largest cities the strata were city districts. In turn, in rural areas, the strata consisted of rural gminas (NUTS5) in subregion (NUTS3), or relatively with a few neighbouring powiats (NUTS4). Overall, there were created 211 strata.
Sampling Method
The sample size for Poland was determined at about 24,700 dwellings. The sample was divided between voivodships proportionally to the square root of the population number of dwellings. In particular, voivodships proportional allocation of the sample of dwellings between strata was carried out. The number of dwellings selected from particular strata was proportional to population number of dwellings in a stratum. In turn, the number of primary sampling units selected from a stratum was a result of dividing the number of dwellings in the sample in a given stratum by the determined for a given class of locality/place number of dwellings to be selected from a primary sampling unit. In the cities with over 500 thousand residents, were selected 5 dwellings from a PSU, in cities with between 100 thousand and 500 thousand residents - 6 dwellings, in towns with between 20 and 100 thousand residents - 7 dwellings from a PSU. In rural areas and in towns with up to 20 thousand residents, 8 dwellings were selected from each PSU. Overall, 3,638 primary sampling units and 24,746 dwellings were selected to the sample.
PSUs were selected with the use of Hartley’a-Rao method. Before sampling, the PSUs were randomly sorted separately in each stratum. Then, a determined number of units was selected with probability proportional to the number of dwellings. Next, in each selected PSU a selection of dwellings was carried out with the use of simple sampling procedure.
When determining the sample size and sampling frame experience from the labour force survey was applied.
Start Date
10-2006
End Date
11-2006
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
15.6 %
Weighting: Method
The applied weighting procedure included the following elements:
1. probability of dwelling selection,
2. level of dwellings’ contacting (addresses),
3. level of completeness of the interviews in households,
4. level of completeness of individual interviews.
The above factors comprised general level of completeness on the level NUTS 2 and were used for preliminary correction of weights.
Then, weights were calibrated separately for each voivodship (NUTS 2) with the use of data from current demographic estimates concerning sex, age (eight 5-year age groups), and place of residence (urban or rural areas).
Weights for 9 larger voivodships were calibrated according to distribution covering simultaneously three variables, i.e. sex x age groups x urban/rural areas. For 7 smaller voivodships were created three marginal distributions in the following form:
- sex x age groups,
- urban/rural areas x age groups,
- urban/rural areas x sex.
Calibrated weights were calculated with the use of the specially created program in the SAS system. The applied calibration procedure was analogous to the one applied in the EU-SIlC.
Resources
-Eurostat: Adult Education Survey. Final Report Poland.
-
EL - Greece
Corresponding National Study
-
Producer
National Statistical Service of Greece
Universe
The target population are individuals aged 25-64 living in private households.
Cross-sectional Data
Achieved Sample Size
3.549
- Cross-sectional
Sampling Design
Probability
- systematic
- stratified
- multi-stage
Primary Sampling Units
Dwellings
Secondary Sampling Units
Households
Stratification Criteria
Prior to sampling, country is stratified by NUT III and degree of urbanization (except Athens and Thessaloniki agglomerations that are divided in 31 and 9 (respectively) strata, depending in socioeconomic criteria). The total number of final strata is 182.
Sampling Method
AES sample was a sub-sample of Labour Force Survey Sample, so it follows the sample design of LFS.
All individuals (of eligible age) in a selected household were interviewed.
Start Date
10-2007
End Date
12-2007
Participation Mandatory
Yes
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
22.9 %
Weighting: Method
At a first stage, a non response correction factor is computed, at primary unit level: Non response correction = number of eligible households/number of households that responded, where number of eligible households = total dwellings in the sample – empty dwellings – (dwellings inhabited by households with all members less than 25 or more than 64).
At a second stage post stratification is carried out: First, totals are computed for 208 “post – stratification cells” (13 NUT II areas x 8 age groups x sex). These totals are computed by weighting sample data (individuals) by design weights x non response correction. Finaly, post stratification correction factors are computed for each individual so that estimations from AES to be consistent with the estimated population values (for the 4th quarter of 2007).
Resources
-Eurostat: Adult Education Survey. Final Report Greece.
-
ES - Spain
Corresponding National Study
-
Producer
National Statistical Institute of Spain
Universe
The target population are individuals aged 25-64, living in Spanish households.
Cross-sectional Data
Achieved Sample Size
18.362
- Cross-sectional
Source Of Sampling Frame
Official Population Register of Spain
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Census enumeration units
Secondary Sampling Units
Individuals
Stratification Criteria
Variables used to stratify: size of the municipalities and main socio-demographics characteristics of the section
The stratums were:
Stratum 0: Municipalities with a population upper than 500.000
Stratum 1: Rest of country towns
Stratum 2: Municipalities no country towns with a population upper than100.000
Stratum 3: Municipalities with a population between 50.000 to 100.000
Stratum 4: Municipalities with a population between 20.000 to 50.000
Stratum 5: Municipalities with a population between 10.000 to 20.000
Stratum 6: Municipalities with a population lower than 10.000"
Sampling Method
In order to be able to make better estimations, we have selected 10 main individuals and 10 substitutes on each section.
In case of ‘ineligible’ main individual, we have used the substitutes individuals with similar age to replace the main individual.
As education is a vble pretty much correlated with education, this kind of substitutions it´s better that not having in mind the age of the individuals.
Start Date
02-2007
End Date
04-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 100 % PAPI
Percentage Of Proxy Interviews
0 %
Weighting: Method
To estimate all the characteristics, ratio estimations with re-weighting techniques has been used. Taken into account the groups of age and sex in the population as auxiliary variables (at regional level).
The following steps has been taken:
1. Obtaining the estimation based on the design, for that, each sampling unit has been multiplied by the inverse of the selection´s probability.
2. Adjusting for unit non-response. The non-response is adjusted by multiplying the later weight by the inverse of the response´s probability within the stratum
3. Ratio Estimator, by using as auxiliary vble the population projection with reference date the survey time. The purpose of the ratio estimator is to improve the one that has been obtained on the previous steps, updating the population that was used when selecting the sample to the population at the data collection time.
4. Finally, reweighting techniques are applied in order to adjust the survey estimations to the population. A reweighting procedure it´s done in order to adjust the estimated population to the distribution of the population by CCAA, age groups and sex. This is the calibrated procces and has been done with the CALMAR macro developed by the INSEE of France. For this calibration next groups have been considered: male and female from 25 to 29, 30 and 34, 35 - 39, 40 - 44, 45 - 49, 50 - 54, 55 – 59 and from 60 to 64 years old. In Ceuta and Melilla (one ccaa) the groups have been different due to the small sample size: males and females from 25 to 34, 35 - 44, 45 – 54 and 55 – 64 years old.
5. The Factor for non formal education activities. It has been selelected ramdomly, at most, 3 of this activities for each of the individuals.
Resources
-Eurostat: Adult Education Survey. Final Report Spain.
-
NL - Netherlands
Corresponding National Study
-
Producer
Statistics Netherlands
Universe
The target population are individuals aged 25-64, living in private households.
Cross-sectional Data
Achieved Sample Size
3.326
- Cross-sectional
Source Of Sampling Frame
Municipal basic registration of population
Sampling Design
Probability
- random
- stratified
- multi-stage
Primary Sampling Units
Municipality
Secondary Sampling Units
Addresses
Stratification Criteria
For the stratification the country is split into 56 strata (areas) that are combinations of 40 Corop areas and 13 interviewing areas (Corop = Commitee for the Co-ordination of a Regional Research Programme).
Sampling Method
The first stage is a stratified systematic probability proportion to size (PPS) sample of municipalities. This means that municipalities are selected within each stratum with probabilities proportional to their numbers of addresses (municipalities in random order).
Subsequently, sample sizes of addresses for the selected municipalities are determined. In the second stage a random sample of addresses is drawn. The sample is self-weighting. Every address has the same inclusion probability 1 / 1000.
One person aged 25-64 per household was interviewed for (added AES questions on) formal education and was consequently asked tot fill in a paper AES questionnaire on non- and informal education and obstacles and/or willingness to participate in education. Potential interviewees in the household were the persons in the “core” of the household, i.e. the person(s) who run(s) the household. The core of the household can consist of one or two persons. If more than one person from a household fell within the field of the survey, one person was randomly selected from the potential interviewees (core of the household).
Start Date
02-2008
End Date
04-2008
Participation Mandatory
No
Type Of Data Source
Survey
Percentage Of Proxy Interviews
22 %
Weighting: Method
The final sample weights of the LFS, which are the starting points of the design weights of the AES were first corrected with the number of individuals in the “core” of the household. The core of the household can consist of one or two persons. If the core of the household consists of one individual, the final sample weight of the LFS is used. If the core of the household consists of two individuals, the weight of the LFS is multiplied by 2 (i.e. the inverse of the response probability within the household).
Using a linear weighing method, a correction factor is calculated. This calibration process adjusted the starting weights so that the estimated totals in the gender-age group (2x4), the education level (6) and the work status (5), fitted the corresponding population values.
The final AES weight is equal to the starting weight times the correction factor.
Resources
Mixed Mode of Data Collection
Eurostat: Adult Education Survey. Final Report Netherlands.
-
EE - Estonia
Corresponding National Study
-
Producer
Statistics Estonia
Universe
The target population are individuals aged 25-64 at the beginning of the survey year and whose address in the Population Register was precise enough to find the person.
Cross-sectional Data
Achieved Sample Size
3.585
- Cross-sectional
Source Of Sampling Frame
Population Register
Sampling Design
Probability
- random
- stratified
- one stage
Primary Sampling Units
Individuals
Secondary Sampling Units
Not used
Stratification Criteria
Stratification is made by age and gender.
Start Date
09-2007
End Date
12-2007
Participation Mandatory
No
Type Of Data Source
Survey
Interview Mode
- 95 % CAPI
- 5 % Self-administered
Percentage Of Proxy Interviews
0 %
Non-response Weight: Method
Eligible units are divided to 7 response homogeneity groups according to place of residence. Non-response correction weights are applied to compensate different response probabilities.
Coverage Adjustment Weight: Method
Variables for calibration are sex and 5-year age group, place of residence. Software package Bascula is used for calibration.
Weighting: Method
Design weights are corrected according to different response probabilities. Individuals are divided into 7 response homogenity group according to place of residence (county and urban/rural). Correction factors are calculated for each RHG as the weighted sum of all eligible individuals divided by weighted sum of responded individuals.
Resources
-Eurostat: Adult Education Survey. Final Report Estonia.