Statistics Estonia

Population and Social Statistics Department

Tatari 51, 10134 Tallinn

The Adult Education Survey (AES) covers adults’ participation in education and training (formal - FED, non-formal - NFE and informal learning - INF). The 2022 AES focuses on people aged 18-69. The reference period for the participation in education and training is the twelve months prior to the interview.

Information available from the AES is grouped around the following topics:

• Participation in formal education, non-formal education and training and informal learning
• Volume of instruction hours
• Characteristics of the learning activities
• Reasons for participating
• Obstacles to participation
• Access to information on learning possibilities and guidance
• Employer financing and costs of learning
• Self-reported language skills

For further information see the 2022 AES legislation (http://ec.europa.eu/eurostat/web/education-and-training/legislation) and the 2022 AES implementation manual (http://ec.europa.eu/eurostat/web/education-and-training/methodology).

• Classification of Learning Activities (CLA, 2016 edition)
• International Standard Classification of Education 2011 (ISCED 2011)
• Classification of Occupations 2008 (ISCO 08)
• Classification of economic activities Rev. 2 (NACE Rev. 2)

AES covers all economic sectors.

Definitions as well as the list of variables covered are available in the 2022 AES implementation manual (http://ec.europa.eu/eurostat/web/education-and-training/methodology).

Individuals, non-formal learning activities.

Individuals aged 18-69 living in private households.

The reference area is the whole country. No parts of the country are excluded.

At national level, the data for reference years 2007, 2011, 2016 and 2022 are available.

Not applicable.

Number, EUR.

Data collection period (fieldwork period) for 2022 AES was 1st July 2022 - 31th December 2022.

At European level:
Basic legal act: Regulation (EU) 2019/1700
Implementing act: Commission Implementing Regulation (EU) 2021/861

At national level:
Not applicable.

Not applicable.

The dissemination of data collected for the purpose of producing official statistics is guided by the requirements provided for in § 32, § 34, § 35 and § 38 of the Official Statistics Act (https://www.riigiteataja.ee/en/eli/522032022003/consolide).

For each respondent, an anonymous ID was generated to ensure confidentiality. No person is identifiable from the data sent to Eurostat or disseminated elsewhere. Only estimates which are based on 20 or more respondents are published in the Statistical Database of Statistics Estonia.

Statistics Estonia released data tables in June 2023.

Not applicable.

All users have been granted equal access to official statistics in Estonia: dissemination dates of official statistics are announced in advance and no user category (incl. Eurostat, state authorities and mass media) is provided access to official statistics before other users. Official statistics are first published in the statistical database. If there is also a news release, it is published simultaneously with data in the statistical database. Official statistics are available on the website at 8:00 a.m. on the date announced in the release calendar.

Every 6 years.

News release about Estonian AES 2022 data: https://www.stat.ee/en/news/two-out-five-adults-would-have-liked-study-more-last-year

No publications except the news release have been published yet. There will be additional blog posts during 2023. 

The aggregated data for key variables can be found in the database of Statistics Estonia https://andmed.stat.ee/en/stat/sotsiaalelu__haridus__taiskasvanute-koolitus__taiskasvanute-haridus.

Not applicable.

Legal persons and organisations can use the microdata for scientific research. The data can be used on a safe centre computer or remotely, depending on the nature of the data and contract conditions. A more detailed description of processing the application and data use conditions can be found in the standard “Procedure for the dissemination of confidential data for scientific purposes": https://www.stat.ee/sites/default/files/2021-07/Procedure%20for%20the%20dissemination%20of%20confidential%20data%20for%20scientific%20purposes_EN.pdf.

There has been no other dissemination.

Not applicable.

The main methodological document is the Eurostat manual. There are no special national reports on methodology, as the Eurostat methodology was followed closely.

Not applicable.

The main documentation on quality management and assessment is the quality report to Eurostat.

To assure the quality of processes and products, Statistics Estonia applies the EFQM Excellence Model, the European Statistics Code of Practice and the Quality Assurance Framework of the European Statistical System (ESS QAF). Statistics Estonia is also guided by the requirements in § 7. “Principles and quality criteria of producing official statistics” of the Official Statistics Act.

Statistics Estonia performs all statistical activities according to an international model (Generic Statistical Business Process Model – GSBPM). According to the GSBPM, the final phase of statistical activities is overall evaluation using information gathered in each phase or sub-process; this information can take many forms, including feedback from users, process metadata, system metrics and suggestions from employees. This information is used to prepare the evaluation report which outlines all the quality problems related to the specific statistical activity and serves as input for improvement actions.

The main users of statistical information are policy makers of national level (Ministry of Education and Research, Ministry of Economic Affairs and Communications, Ministry of Social Affairs). The data might also interest the Estonian Qualifications Authority, different research institutions and university researchers.

Internationally, the main users of the data are the institutions of the European Union. In addition, through Eurostat, researchers around the world can also apply for access to use the data.

Since 1996, Statistics Estonia has conducted reputation and user satisfaction surveys.

All results are available on the website at https://www.stat.ee/user-surveys.

The satisfaction of users specifically with AES has not been analysed.

Dataset covers all requested variables in AES 2022 legislation.

Not applicable.

Coefficients of variation were calculated in R. The accuracy varies. Estimates based on a high number of observations are more accurate than those based on less observations.

Indicators for which a precision threshold is provided in the AES Regulation are of good quality.

The sampling error for 2022 AES is fairly low. Stratified systematic sampling was used to create the sample. The estimates for the sampling error were calculated using the R package survey with the final (calibrated) weights.

The coefficients of variation, standard errors, and confidence intervals were calculated using the R package survey with the final (calibrated) weights.

See table 13.2.1 “Sampling errors - indicators for 2022 AES key statistics” in annex “EE - QR tables 2022 AES (excel)”.

Non-sampling errors are covered by items 13.3.1 - 13.3.5 below.

Sampling frame: List of permanent residents of Estonia aged 18-69 from the statistical population register, age calculated as of 01.07.2022. Date of extraction 01.05.2022.

Over-coverage error was 1.6%. 

See table 13.3.1.1 “Over-coverage - rate” in annex “EE - QR tables 2022 AES (excel)”.

Not applicable.

In order to reduce the measurement errors, the AES questionnaire was tested by experts before data collection, and an interviewer manual was compiled to explain specific terms and questions in particular. AES was a stand-alone survey and proxy answers were not allowed.

See items 13.3.3.1 - 13.3.3.2 below

See table 13.3.3.1 “Unit non-response - rate” in annex “EE - QR tables 2022 AES (excel)”.

Advance notifications about the survey were sent to the respondents. These letters included information about the time period, when the interviewer will make a phone call. If the respondent was not reachable by phone, the interviewer made a home visit and made a personal interview.

See table 13.3.3.2 “Item non-response rate” in annex “EE - QR tables 2022 AES (excel)”.

Item non-response is very low. Only 3 variables have been measured with a non-response rate between 5-10%, and a majority of variables have recorded a non-response rate of less than 1%.

Questions for HHINCOME were very sensitive as 37% needed imputing.

Computer-based data collection is used, which helps to avoid encoding and typing errors. Data is checked in three stages: initial input control during the interview on a laptop, secondary control of freshly received data in the office, and finally, data cleaning. Last validation was Eurostat CONVAL and STRUVAl controls in Edamis.

R was used to code the initial data according to the codes in the AES manual.

Not applicable.

The reference period for the 2022 AES was 12 months prior to the interview, that took place in the following period: from 1st July 2022 until 31th December 2022. Data were published nationally within 6 months after the end of the fieldwork period.

Data was published in June 2023.

Data was published in June 2023.

The microdata was sent to Eurostat within six months of the end of the national fieldwork period. The data was published as scheduled in the release calendar.

See table 14.2 “Project phases - dates” in annex “EE - QR tables 2022 AES (excel)”.

Not applicable.

The data is comparable to the data from other European Union countries. 

See table 15.1 “Deviations from 2022 AES concepts and definitions” in annex “EE - QR tables 2022 AES (excel)”.

Not applicable.

See table 15.2 “Comparability - over time” in annex “EE - QR tables 2022 AES (excel)”.

Starting from the 2022 survey, the non-formal education activities were selected randomly. Previously, the individual had the opportunity to choose which trainings they provided information about.

In 2011, the results of the adult education survey showed that the percentage of people participating in non-formal education in the previous 12 months was the highest in timeline. However, this conclusion does not align with the results from the Estonian Labour Force Survey, which has shown a steady increase in lifelong learning participation without a noticeable change around 2011. The Labour Force Survey's data is about training events that occurred over a 4-week period, but the trends can still be compared.

The 2011 survey was different from others in that it was not solely conducted by the Statistics Estonia, but was carried out in collaboration with a subcontractor. Additionally, the response rate that year was lower than usual. Consequently, there may be more deviations in these data than usual (the non-sampling error is greater). To achieve a more consistent overview over time, it may be advisable in some cases to exclude the 2011 data from the analysis.

Not applicable.

AES population structure:

The size of the population derived from the survey data does not match the total population of the same age range from registry because those in institutions are excluded. AES 2022 was calibrated to the population count as of 01.01.2022 (minus those in institutions).

AES and LFS can be compared, but since they have different age distributions and we don't use age*education for calibration, there are naturally differences.

See table 15.3 “Coherence - cross-domain” in annex “EE - QR tables 2022 AES (excel)”.

Not applicable.

Not applicable.

AES results for a given data collection round are based on the same microdata and results are calculated using the same estimation methods, therefore the data are internally coherent.

The staff involved in administrating the survey (project management, sample design, questionnaire development, training the interviewers, weighting, data processing, data analysis and quality report) ~ 375 full-time equivalent (FTE) working days.

The total fieldwork time spent conducting CATI interviews, which involves direct contact with respondents, was approximately 129 full-time equivalent (FTE) working days. However, the time interviewers spent on training, preparation, attempted calls, and other indirect tasks, was not measured.

The average time for answering the questionnaire:

• CAWI: 32 minutes (883 interviews)
• CATI: 18 minutes (3477 interviews)

The average time for answering the questionnaire when a person had participated in at least 2 non-formal activities( NFENUM>=2):

• CAWI: 41 minutes (353 interviews)
• CATI: 27 minutes (716 interviews)

Not applicable.

Not applicable.

Not applicable.

Survey data: The data is based mainly on national stand-alone survey. Participation in the survey was voluntary.

Sampling frame: List of permanent residents of Estonia aged 18-69 from the statistical population register, age calculated as of 01.07.2022. Total population is 873,000 objects.

Sampling design: Stratified random sampling by age and sex. Gross sample 6800, net sample 4360.

Administrative data: There are no substantive data used from registries. However, background characteristics are assigned based on the population register data: gender, year of birth, age, county of residence, urbanity of the place of residence.

See table 18.1 “Source data” in annex “EE - QR tables 2022 AES (excel)”.

Every 6 years.

AES in Estonia was a stand-alone survey. Data was collected from individuals. The methods used to collect the data included a telephone interview (CATI) and a web interview (CAWI). The interviews were conducted by Statistics Estonia's telephone interviewers who had relevant training. The Survey Fieldwork Information System was used to manage and monitor data collection.

See also table 18.1 “Source data” in annex “EE - QR tables 2022 AES (excel)”.

The data was collected with the official statistical questionnaire "TÄISKASVANUTE KOOLITUSE UURING 2022," available in both Estonian and Russian. Please see the questionnaire in the annex titled 'EE_manual_AES2022_est' and the guidelines for interviewers in the annex 'EE_questionnaire_manual_AES2022_est'.

The validation process consists of arithmetic and quality checks, including comparison with other data. Before data dissemination, the internal coherence of the data is checked.

There were many logical checks and controls in the survey, which did not allow to enter impossible answers. Data was also re-checked after the collection period.

Imputation:

In the case of missing or unreliable data, estimate imputation based on established regulations is used. For household income with high non-response we used the following method: Respondents were asked to either provide their exact household income or indicate the interval to which their household income belongs. In case of both being unknown the interval is first imputed using k-nearest neighbour imputation using the VIM package in R (the number of nearest neighbours was set as 5 and household size and type and the respondent's age, sex, county + Tallinn, and educational attainment were used as distance variables). Once every respondent was supplied with an income interval the precise household income was imputed using hot-deck imputation.

Weights:

After the data processing stage, each respondent is assigned a weight, i.e., it is determined how many people their responses represent. Weights allow the results obtained during analysis to be extrapolated to the general population.

Weights are calculated in three stages: Calculation of design weights, Compensation for loss, Calibration. Calibration is based on gender and age group (five years age groups), nationality (since 2005) (Estonians, non-Estonians), three-tier education level (since 2022), the person's county of residence (+Tallinn), and the degree of urbanization (rural or urban). The basis is the known distribution of Estonian residents by gender, age group, and county as of January 1 of the survey year, based on demographic data.

After calculating calibration weights, a general expansion factor is found for each responding person, which is the ratio of the total population size to the number of respondents. To get the final weight for the person, this expansion factor is multiplied by the previously calculated weights.

Calculated variables:

Variables and statistical units which were not collected but which are necessary for producing the output are calculated. New variables are calculated by applying arithmetic conversion to already existing variables. This may be done repeatedly, the derived variable may, in turn, be based on previously derived new variables. These variables were: HHTYPE, LANGUSED, HHINCOME, NFEACTWEIGHT_2, NFEACTWEIGHT_5.

See table 18.5.1 “Imputation - rate” in annex “EE - QR tables 2022 AES (excel)”.

Not applicable.

Not applicable.

None.