Istat - Italian National Statistical Institute

SWB - Division for integrated system for labour, education and training

Via Cesare Balbo, 16 - 00184 Rome Italy

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)
- International Standard Classification of Education: Fields of Education and Training 2013 (ISCED-F 2013)
- 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 (as in 2016, the individuals aged 70-74 are also covered in the national dataset).

Italy

The data are available from 2012 (2011 AES).

Not applicable.

Number, EUR.

Fieldwork period for 2022 AES: September 15, 2022 - January 15, 2023.

At European level:
Basic legal act: Regulation (EU) 2019/1700
Commission Delegated Regulation (EU) 2021/859
Commission Implementing Regulation (EU) 2021/861

At national level:
The process is included in the National Statistical Programme (PSN) that is annually approved by Decree of the President of Republic (DPR) on the basis of the Law n.125 of 30 October 2013. In particular, the survey is included in the PSN in force approved with DPR of 11 July 2023 with code IST-02816.

Not applicable.

Several national legal acts guarantee the confidentiality of data requested for statistical purposes. In Italy, according to art. 9, paragraph 1 of the Legislative Decree n. 322 of 1989 (concerning the statistical system), statistical data cannot be disseminated but in aggregated form, in order to make it impossible to identify the person to whom the information relates. The data collected can only be used for statistical purposes.

Official statistics must also safeguard the rights, basic freedoms, and dignity of respondents, in particular with regard to the right of confidentiality and personal identity.

Istat assures the protection of personal data according to the General Data Protection Regulation (Regulation (EU) 2016/679 of the European Parliament and of the Council of 27 April 2016 on the protection of natural persons with regard to the processing of personal data and on the free movement of such data, repealing Directive 95/46/EC) and, as national legislation, Italian Data Protection Code (Legislative Decree no. 196/2003) and Code of conduct and professional practice applying to the processing of personal data for statistical and scientific research purposes within the framework of the national statistical system.

In order to make statistical secrecy and protection of personal data effective, Istat is currently taking appropriate organizational, logistical, methodological and statistical measures in accordance with internationally established standards.

Moreover, Legislative Decree n. 322 of 1989, art. 6 and 6 bis provides that the exchange of microdata and personal data within the National Statistical System (Sistan) is possible if it is necessary to fulfil requirements provided by the National Statistical Programme.

Finally, in implementation of art. 5-ter of the legislative decree 14 March 2013, no. 33, the new "Guidelines for the access for scientific purposes to the elementary data of the National Statistical System" establish the conditions under which the bodies and offices of the National Statistical System can allow researchers to access their own elementary data for scientific purposes.

Before data transmission to Eurostat, all respondent identifiers were removed from the collected data (anonymization).

At the end of each year Istat defines and publishes a release calendar for the following year for press releases. The release calendar is then updated weekly to include all the other Istat releases.

The release calendar is available on the website:

https://www.istat.it/en/information-and-services/journalists/press-releases-/press-calendar

According to its mission, Istat disseminates statistical information in order to make it accessible and usable to everyone and to remove any barriers to the use of data. All data releases are posted on Istat website on the basis, as regards press releases with short-term data and annual data of strong interest for the country, of an annual release calendar set and published in the December preceding the reference year. In general, time series are available on Istat corporate data warehouse and users can choose information according to their needs, building customised tables or downloading pre-packaged datasets. AES data are not published on Istat data warehouse since they will be published on Eurostat database. Data are always accompanied by meta-information - methodologies, classifications, definitions. Microdata files are released free of charge and in compliance with the principle of statistical secrecy and data protection. Books, press releases, datasets and infographics are also available on Istat web site; moreover main contents are disseminated through Istat Official Twitter account and other social networks. All Istat information are available free of charge and data are reusable providing the source. As for previous editions, AES data will be mainly disseminated through an ad-hoc statistical report planned for the end of 2023.

Every 6 years.

Adult education and training. Year 2017 - Anno 2017
The usage of Italian language, dialects and other languages in Italy. Year 2012 - Anno 2012
Adults' Participation in Learning Activities. Year 2012 - Anno 2012
At national level, a first data release is currently scheduled in December 2023 (Statistiche report).

Not applicable.

AES Data will be released on the Eurostat website.

Not applicable.

Not applicable.

Eurostat data warehouse: http://ec.europa.eu/eurostat/data/database.

Not applicable.

For 2016 edition documentation on methodology is available at https://circabc.europa.eu/ui/group/d14c857a-601d-438a-b878-4b4cebd0e10f/library/4581648f-0ce9-4b59-beaf-e07d050c7de6/details 

With the first release of 2022 data Istat will also release a report on the Italian methodology.

Not applicable.

The Istat Information system on quality of statistical production processes SIQual contains information on the execution on Istat statistical production processes and on activities developed to guarantee quality of the produced statistical information. For details: https://siqual.istat.it/SIQual/visualizza.do?id=8889001&language=UK.

Since the 90s Istat has adopted a systematic approach to ensure the quality of statistical information and of its services to the community.

With the aim of strengthening the commitment to quality, in 2020 Istat set up the Quality Committee, for overseeing all quality initiatives in the Statistical Institute. In addition the role of Quality Manager was formally established.

In 2021 a new quality policy for statistical production was adopted. It is consistent with the European quality framework developed by Eurostat, and transposes its main principles and definitions.

For details: https://www.istat.it/en/organisation-and-activity/institutional-activities/quality-commitment.

The process was submitted to Quality statistical auditing for the 2012 edition (2011 AES).

AES data are used to get a picture of life-long learning. In particular:

• Participation in education and training by type, characteristics of the activity (field, distance learning, etc.)
• Access to information on learning possibilities and guidance 
• Reason, use and outcomes of FED and NFE 
• Share of job-related or employer-sponsored NFE
• Volume of instruction hours for FED and NFE
• Cost of learning for NFE
• Obstacles to participation in education and training

The main users are:

• Policy makers at European level (e.g. European Commission, European Parliament, other European agencies)
• Policy makers at national level (e.g. ministries)
• Social actors (e.g. employers' associations, trade unions)
• Media
• Researchers, students
• Enterprises: for own market research activities or for consultancy services in the information sector
• International organisations (OECD, UN)

Istat is constantly interested in understanding who the users of the statistics it produces are, what the information needs are, whether they match production and if the statistics produced satisfy users. To this aim, together with the analysis of user requests received through the Web Contact Center service, tools for direct consultation were developed, such as the annual online survey of customer satisfaction and indirect tools such as analysis of accesses and of users' browsing paths on the web site.

All variables required by the Regulation were transmitted.

Not applicable.

In order to reduce the amount and the impact of non-sampling error several quality control actions have been carried out during the survey.

The first step aimed at facilitating field operations has been the sending of an informative letter to all the sampled individuals informing them of their participation in the survey. The letter briefly described objectives and methods of the survey, with particular reference to its relevance for the purposes of national and European policies, as well as the regulatory aspects governing it (mandatory participation, sensitive questions, etc.). In fact, it is known that the response rate of those who are notified institutionally is always significantly higher than those who are not.

CAWI respondents received also a first reminder after 30 days from start of the data collection, followed by several others during the data collection period.

For the CATI sample, a toll-free number was available where a team of trained operators reassures callers about the authenticity of the survey and confirms the accreditation of the interviewer.

For the CAWI sample, another group of more qualified telephone operators provided technical support to callers (questionnaire access, credential retrieval, etc.) and forwarded the request to Istat experts when thematic support was required. Thematic support was provided to respondents via email or telephone. Most of the thematic questions concerned the codification of some variables (jobisco, locnace, fedfield and nfefield).

Particular attention was paid to the recruitment and training of the CATI-interviewers. They were only involved in the survey after attending the training sessions scheduled before the survey began.

CAWI and CATI software also allowed to have indicators continuously updated on the progress of the survey (number of contacts, number of complete interviews, number of refusals and reason for refusal, duration of the interview, etc.) both at territorial and interviewer level (each selected individual was associated with a single interviewer).

Finally during the data collection, continuous monitoring was carried out on completed interviews through a weekly download of the data and their check to detect errors and inconsistencies. Based on these checks, two weeks after the start of the survey, a new training session was carried out for the CATI interviewers in order to reduce systematic errors.

Non-sampling errors were treated through data editing and imputation procedures.

For sampling error see 13.2.

Sampling error was estimated from the sample data through the linearization formula of the sampling variance (Re-genesees software) of the requested indicators.

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

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

Calibration estimators (that are partially useful to avoid under-coverage errors) were used to adjust for coverage errors.

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

Not applicable.

Rules embedded in the electronic questionnaire, together with other preventive and monitoring actions like the training of interviewers, reduced the amount of measurement errors in collected data, however editing and imputation procedures were used to identify and correct remaining errors.

The quantitative variables HHINCOME, NFEPAIDVAL1, NFEPAIDVAL2, NFENBHOURS1, NFENBHOURS2 and FEDNBHOURS had the most significant impact on measurement errors. For these variables, probabilistic procedures have been employed to reduce measurement error (see 18.5).

For the other variables data imputation has been very limited and deterministic procedures were used.

The questionnaire did not allow for item non-response except for the two variables GENHEALTH and GALI because of their sensitive nature. Otherwise, all questions were considered mandatory.

For some questions that were considered more complex, the respondent was given multiple alternative answers.

Despite their loss of precision, they prevented in fact item non-response and the early interruption of the interview.

These include income and expenditure variables, the year of specific events, and the classification of same variables (Jobisco, Locnace, Fedfield, and Nfefield).

This made it necessary, in particular for CAWI respondents, an a posteriori recoding based on the information provided by the respondent during the interview.

Similarly, the items "other" have been reclassified where possible.

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

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

An electronic questionnaire was used for both the CAWI and the CATI surveys, which limited the post-data collection data processing. Therefore it is considered not to have introduced substantial errors (processing error not relevant).

Not applicable.

As required by the Regulation data has been transmitted within six months from the end of the national data collection period.

At national level, the data dissemination is scheduled within twelve months from the end of the national data collection period.

Not applicable.

Scheduled time lag for final result was approx. 181 days, Data were transmitted within 173 days.

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

As data were transmitted 8 days before the deadline, punctual.

Not applicable.

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

In Section N of the questionnaire, all COVID variables recommended by Eurostat in February 2022 were introduced.

In the same section N, other questions about online courses (pros and cons, tools and home environment) were also added.

Not applicable.

Six years after the previous edition, the survey design was substantially modified.

The sampling design, the collection technique and the questionnaire have been modified to better meet Eurostat’s recommendations, in addition to the amendments set in the Regulation. 

This has affected the comparability of results over time for all variables and in particular for NFE and INF participation and for some multiple-choice variables (FEDREASON, FEDOUTCOME, NFEREASON1, NFEOUTCOME1, NFEREASON2, NFEOUTCOME2, DIFFEREASON2). In fact for all the multiple-choice variables the "single checklist" was preferred instead of the "yes/no checklist" used in the previous edition both to better meet Eurostat’s recommendations and to simplify the auto-filling of the questionnaire especially for CAWI respondents.

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

Not applicable.

See table 15.3 “Coherence - cross-domain” in annex “IT - 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 CAWI survey, as expected, reduced economic costs and offered the respondent greater flexibility for compilation.

In order to reduce the burden on respondents due to the length of the questionnaire and the complexity of some questions, action has been taken either in the formulation of questions (by simplifying the language, inserting explanatory pop-ups, etc.) and by reviewing the automatic checks included in the electronic questionnaire (in order not to discourage the respondent with too many warnings).

The CAWI response rate is close to that expected (about 30%).

During the fieldwork phase, the CAWI response team (contact center operators) responded to several requests, solving both technical problems and thematic doubts. At the same time, the a posteriori interventions, especially for the post-coding of some variables and reallocating the items 'other', have been more significant than in the CATI survey.

However, the quality of the data collected in CAWI is absolutely satisfactory and comparable to the CATI data.

Not applicable.

Not applicable.

Not applicable.

Data derive from a sample survey. The sample design was integrated with the sample of the Permanent Population Census (Master sample). In the specific case, the sample units for the AES were selected as a sub-sample of the set of individual respondents to the Master Sample for 2021 (this deriving from a two staged sample design, municipalities-households). A stratified sample design has been defined, where the strata are the cross-classification of the following variables: NUTS2, Italian regions (21 regions); municipal type (metropolitan cities, municipality in the metropolitan area, other municipalities subdivided in >10000 inhabitants, 10000-50000 inhabitants, >50000 inhabitants); citizenship (Italian, foreigner); 7 age classes (18-24, 25-34, 35-44, 45-54, 55-59, 60-64, 65-69) and sex. The overall sample size and its allocation among strata was defined to satisfy the requested constraints about the sampling errors of the main estimates. A total theoretical sample of 40000 individuals was defined, to satisfy also national estimation needs.

Units were selected from the set of eligible respondents to the Master Sample with equal probability in each stratum. The sample selection was carried out separately for the CAWI and the CATI samples, according to the availability of a telephone contact: for the CAWI sample (selected from the set of units without a telephone contact) an oversampling was done, based on response rates derived from a previous similar survey; for the CATI sample (selected from the set of units with a telephone contact) quadruplets were selected for possible replacements.

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

Every 6 years.

Computer Assisted Telephone Interview (CATI)

Computer Assisted Web Interview (CAWI)

External data sources are used to verify and validate the collected data (administrative data, LBS data, EU-SILC data and other national surveys data). A general and rough comparison has also been made with the previous AES edition data.

A description of data editing and data estimation and imputation is provided.

Within the survey, different editing and imputation methods were used for different variables.

For the variables related to education expenditure, and for household income, outliers and influential errors were identified.

To identify outliers and influential errors, a specific latent class model was employed for expenditures. The ‘number of hours’ of the course was used as a covariate. Imputation was performed replacing the incorrect or missing value with the value predicted by the model. The R package Selemix has been used.

For the household income both graphical analysis of the empirical distribution and consideration of the specific characteristics of the observed phenomenon were used to determine an acceptance range. Values falling within the first two and last two percentiles of the distribution were considered outliers. In order to identify incorrect outliers, several covariates, significant for analyzing household incomes, were taken into account. Then, multiple imputation was performed using the IVEware software's IMPUTE module, employing a multivariate sequential regression approach. Five imputations were generated for each record to be imputed. The final imputed value was obtained by taking the mean of these five values.

For the other variables, such as the number of hours for FED and NFE courses, data editing was carried out in the R environment using validate and validatetools packages to check data with respect to the rules of the questionnaire and the logical inconsistencies of combinations of values. Imputation of the variables affected by errors or missing values was done with the R-package VIM, implementing hot-deck and kNN imputation methods.

Final weights were calculated in three steps. 

First step

A direct sampling weight for each respondent unit was calculated considering the selection of the sample from the set of the respondents to the population census and also the weights of the census respondents to the whole population (the first weight was multiplied by the second).

Second step

An adjustment of direct weights was performed for correcting non-response, separately for the CAWI and the CATI samples. For the CAWI sample, for which the theoretical sample corresponds to the selected sample, this was obtained through the response propensity method, exploiting the auxiliary variables known for respondent and non-respondent units from the census data set. The auxiliary variables used in the response model are: 2 classes of citizenship (Italian, not Italian), 5 age groups (18-30, 31-40, 41-50, 51-60, 61+), 3 classes of educational degree (ISCED 0-2; ISCED 3-4; ISCED 5-8), work-employment position (employees, self-employed workers, students, non-employed and non-students); 5 main geographic areas. Weight adjustments are obtained using the inverse of the response propensity estimated by the gradient boosting model.

For the CATI sample a post-stratification with respect to the overall selected sample (base and replacement units) was calculated on the same auxiliary variables as for the CAWI sample. 

The output of this second phase is a base weight, given by the product of the direct weight by the nonresponse factor.

Third step

A final calibration with known totals referred to the time of the survey was performed to correct the base weights, pooling the two samples. The calibration variables used to calculate the final weights are sex, type of municipality and educational level. For each of these variables the known totals were calculated at the regional level, cross classified as follows: 1. individuals for sex and 7 age groups (18-24, 25-34, 35-44, 45-54, 55-59, 60-64, 65-69); 2. individuals for municipal type; 3. Individuals for 3 educational degree (ISCED 0-2; ISCED 3-4; ISCED 5-8) and 3 age groups (18-34, 35-54, 55-69). The totals for the educational levels derive from the population register, the other totals derive from demographic source.

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

Not applicable.

Not applicable.

None.