Education at a Glance 2023 Sources, Methodologies and Technical Notes

1. Sources, methodologies and technical notes for Chapter A

List of indicators

• Indicator A1. To what level have adults studied?

• Indicator A2. Transition from education to work: Where are today’s youth?

• Indicator A3. How does educational attainment affect participation in the labour market?

• Indicator A4. What are the earnings advantages from education?

• Indicator A5. What are the incentives to invest in education?

• Indicator A6. How are social outcomes related to education?

• Indicator A7. To what extent do adults participate in education and training?

Description

This document is intended to provide guidance as to the methodology used during the data collection for each indicator, the references to the sources and the specific notes for each country. For general information on methodology, please refer to the OECD Handbook for Internationally Comparative Education Statistics: Concepts, Standards, Definitions and Classifications (OECD, 2018[1]).

Table 1.1. Sources, methodologies and technical notes for Chapter A

Tables X3.A1. Sources, methodologies and technical notes for Indicator A1
Table X3.A1.1	Codes from ISCED 2011 used for describing educational levels
Table X3.A1.2	National data collection sources and reliability thresholds for the NEAC questionnaire
Table X3.A1.3	Educational attainment and associated standard errors, by age group and gender (2022)
Tables X3.A2. Sources, methodologies and technical notes for Indicator A2
Table X3.A2.1	National data collection sources and reliability thresholds for the TRANS questionnaire
Table X3.A2.2	Percentage of young adults in education/not in education and associated standard errors, by age group, gender and work status (2022)
Tables X3.A3. Sources, methodologies and technical notes for Indicator A3
Table X3.A3.1	Employment rates and associated standard errors, by age group, educational attainment and gender (2022)
Table X3.A3.2	Unemployment rates and associated standard errors, by age group, educational attainment and gender (2022)
Tables X3.A4. Sources, methodologies and technical notes for Indicator A4
Table X3.A4.1	National data collection sources and reliability thresholds for the Earnings questionnaire
Table X3.A4.2	Data coverage for the Earnings questionnaire
Table X3.A4.3	Treatment of individuals with zero or negative earnings in the Earnings questionnaire
Table X3.A4.4	Actual earnings of full- and part-time workers, by age group, educational attainment and gender (2021)
Table X3.A4.5	Actual earnings of full-time full-year workers, by age group, educational attainment and gender (2021)
Table X3.A4.6	Coverage of individuals with no earnings from work

Statlink

Indicator A1. To what level have adults studied?

Methodology

The educational attainment profiles for most countries are based on the percentage of the population that has completed a specific level of education. The International Standard Classification of Education (ISCED) is used to define the levels of education.

In Education at a Glance (EAG), ISCED 2011 is used to classify the levels of education. Unless data using national codes according to ISCED 2011 have been provided by countries, trend data on educational attainment are only available for the three aggregated levels of education. The linkage between ISCED-97 and ISCED 2011 for the three aggregated levels of education are the following:

• Below upper secondary education includes either the codes 0/1/2 if data are classified using ISCED 2011 or codes 0/1/2/3C short if data are classified using ISCED-97

• Upper secondary or post-secondary non-tertiary education includes either the codes 3/4 if data are classified using ISCED 2011 or codes 3/3C long/4 if data are classified using ISCED-97

• Tertiary education includes either the codes 5/6/7/8 if data are classified using ISCED 2011 or codes 5A/5B/6 if data are classified using ISCED-97

Table X3.A1.1 (see StatLink under Table 1.1) shows the educational attainment and ISCED mappings/codes for each country. It presents the national codes according to ISCED 2011 and shows the codes included in each level of education.

Notes on new methodology of the European Union Labour Force Survey (EU-LFS)

For European countries providing data from the EU-LFS there was a new LFS legislation in 2021 and Eurostat flag all 2021 LFS data with “b”, i.e. break in series, in particular for labour-market outcomes indicators. Information on the impacts of the new methodology is not available by the time Education at a Glance 2022 was released. For general information and country specific notes on the changes, please see the webpage from Eurostat.

Source

The OECD Network on labour market, economic and social outcomes of learning (LSO Network) compiles the data from the national Labour Force Surveys (LFS) for Tables A1.1, A1.2 and A1.3 on educational attainment. Data for table A1.4 on recent graduates were extracted from EU-LFS for all countries participating in this survey by Eurostat.

• For tables with trend data, the European Union LFS (EU-LFS) has been used for Denmark, Estonia and Latvia for the years 2000, 2005 and 2010.

• Data for China, India, Indonesia and Saudi Arabia were taken from the ILO database, https://www.ilo.org/global/statistics-and-databases/lang--en/index.htm (accessed on 22 May 2023).

Specific reliability and confidentiality thresholds by country have been applied. The reliability and confidentiality thresholds are either applied to unweighted data (respondents) or weighted data (population):

• Reliability thresholds refer to sample limits to the statistical precision of the indicator that implies that data are either not published or flagged as having a reduced reliability. Reliability thresholds have been applied to the denominator of the indicator. A minimum threshold of 30 respondents for unweighted data have been applied to the denominator.

• Data may be omitted due to confidentiality reasons. The respective confidentiality threshold has been applied to the numerator of the indicator. A minimum threshold of 3 respondents for unweighted data have been applied to the numerator.

Table X3.A1.2 (see StatLink under Table 1.1) summarises the metadata by country.

Notes on specific countries

Austria

ISCED 0-2 includes ISCED 3c short from 2006 onwards (measured as successful attainment of intermediate technical and vocational school shorter than two years), 2004 and 2005 ISCED 3c short covers also intermediate technical and vocational schools of a duration of two years. For 2004 and 2005 ISCED 3c short is therefore reported as upper level.

Due to the reclassification of a programme spanning levels data published from Education at a Glance 2015 on are not directly comparable with data published on previous editions of Education at a Glance. The qualification acquired upon successful completion of higher technical and vocational colleges is allocated in ISCED 2011 to ISCED level 5; under ISCED 1997 the same qualification was reported on ISCED level 4 but earmarked as equivalent to tertiary education.

Belgium

A break in time series occurred in 2017 as the Belgian LFS has undergone a major reform in 2017 with the introduction of a 2(2)2 panel design, the introduction of mixed mode data collection, the introduction of the wave approach and the change of the calibration method.

Canada

The Canadian Labour Force Survey does not allow for a clear delineation of attainment at ISCED 4 and at ISCED 5; as a result, some credentials that should be classified as ISCED 4 cannot be identified and are therefore included in ISCED 5. Thus, the proportion of the population with tertiary education ISCED level 5 is inflated. It is also not possible to single out university credentials above bachelor’s (ISCED 6) and therefore doctoral levels (ISCED 8) are included in Masters or equivalent (ISCED 7), resulting in the overstating of Masters or equivalent (ISCED 7).

Statistics Canada has an established history of applying a standard revision to LFS estimates following the release of final population estimates from each census. This standard revision process ensures the continued accuracy and quality of LFS information by ensuring that survey estimates reflect the size and composition of Canadian society. Resulting changes to recent and historical LFS data are relatively minor and have little impact on trends in key labour market indicators, such as employment, unemployment, and participation rates. Until December 2020, labour force estimates were based on population counts from the 2011 Census of Population. As of January 2021, the estimates have been adjusted to reflect population counts from the 2016 Census, and revisions back to 2006 have also been made.https://www150.statcan.gc.ca/n1/en/catalogue/71F0031X

Chile

The National Socio-Economic Characterization Survey (CASEN) does not allow determine the orientation of upper secondary educational programmes (ISCED 344/354) for those with incomplete tertiary education. Observations in this situation have been classified as general upper secondary education (ISCED 344).

Denmark

A break in time series occurred in 2007 as the survey was changed and expanded considerably, by expanding the quarterly sample size from around 20,000 to 40,532 in order to reduce sampling errors of survey results. Furthermore, the rotation pattern was changed from three to four waves, and the data collection process which Statistics Denmark had been in charge of so far was outsourced. The changes in 2007 resulted in a break in series both on detailed sub-groups. As a result of this one should be aware of this when comparing results before and after the break.

In 2019 the weighting scheme was changed in the Danish LFS as the Employees Statistics, information on transfer income from the eIncome register and information on ownership of sole proprietorships and partnership enterprises from Statistics Denmark’s business register had become sufficiently up to date to be applied in LFS’s raising model. Altogether, over 90 per cent of the respondents in the LFS are registered in one or more of the mentioned relevant registers. By ensuring that the number of persons in employment – as well as other labour market statuses in the LFS – better match the number they represent in the population, different response rates of these different groups are adjusted for. Thereby bias is reduced.

The Employees Statistics goes back to 2008, so implementing this new raising model has created a break in time series between 2007 and 2008.

Estonia

In tables with trend data, the European Union Labour Force Survey (EU-LFS) data were used for all years.

Finland

At the beginning of 2021, the data content, data collection and estimation method of the Finnish Labour Force Survey were revised. The most important content changes concerned working time data and the definition of employed person. The data collection was revised by introducing a web questionnaire as a new data collection method alongside computer-assisted telephone interviews and face-to-face interviews. The revised estimation method utilises more widely different registers, such as Statistics Finland's Register of Completed Education and Degrees. For example, the Register of Completed Education and Degrees is used to correct the skewness of the collected Labour Force Survey data according to the level of education more precisely than before.

Revisions of the Finnish Labour Force Survey were not related to the COVID-19 pandemic. The revisions were general development of the Labour Force Survey and due to the revisions of the Integrated European Social Statistics. The statistics produced with the old estimation method (for data prior to 2021) are not comparable with the statistics produced with the new estimation method.

France

Variables on education attainment changed in 2003 and 2013 (more accurate). From 2003, age is reported at survey time instead of the end of the year. In 2013, the questions on educational attainment have been simplified and the process became more interactive (impact of about +2 percentage points of the proportion of 25-64 with ISCED 3-8). Furthermore, in 2013 a break in the series of unemployment rates occurred. From 2015, data cover overseas departments.

Greece

The adoption of the new Framework Regulation of Social Statistics (ΕU 2019/1700, Integrated European Social Statistics) and the Implementing Regulation (ΕΕ) 2019/2240 for the Labour Force Survey, introduced important changes in the survey methodology. The main changes concern: i. Data collection (general use of computer-aided interviews); ii. The formulation of questions related to the employment status of respondents during the reference week (due to the adoption of a model questionnaire prepared by Eurostat); iii. The computation of the weighting factors; iv. The definition of employment and, in particular, the treatment of person reporting having a job during but not working even for one hour.

Started from January 2018 Labour Force Survey questionnaire was expanded and changed: 1. Starting from 2018 the absent from work workers are asked how many hours they usually work; 2. Involuntary part-time workers' definition has changed:

• Until 2017: Employees and co-operative members usually working less than 35 hours per week (incl. employees owners of Ltd companies), because they sought but did not find full-time or additional work.

• As of 2018: Employees and co-operative members (excl. employees owners of Ltd companies) usually working less than 35 hours per week, who are interested in working more than 35 hours and actively sought to work more hours in the last four. Excl. persons living outside localities (Bedouins in the South) or in institutions (permanent samples).

Italy

During 2021, the Labour Force Survey underwent two major changes:

1. the introduction, from 1 January 2021, of the European Regulation (EU) 2019/1700, which entailed changes in definitions and questionnaire in order to improve the degree of harmonisation of the statistics produced by the EU countries, to optimize the information already collected, to meet current needs for information and to explore specific issues;

2. the adaptation to the new population estimates of individuals and families derived from the Permanent Census of Population, which aims to improve the quality of demographic statistics in terms of consistency, structural composition and number of events observed, also in order to ensure that the results of the sample surveys are representative of the resident population.

For these reasons, a process of back-recalculation has been conducted and new time series are now available, starting from 2004 for the main indicators and from 2018 for more detailed series.

The questions concerning the education attainment were revised without affecting the education attainment data on below upper secondary, upper secondary and tertiary aggregates, although some minor impact may have occurred at more disaggregated level or on specific age-classes. Some slight impact may also occur when the participation is analysed separately for formal and non-formal education, because the changes introduced in the 2021 questionnaire.

Moreover, some indicators are still under study and among them there is the information on usual working hours. Considering that data collected during 2020 and 2021 were strongly affected by the pandemic, for this topic it is necessary to acquire further data to assess the seasonality of the estimates from the new survey and to evaluate the need to proceed with the reconstruction of the entire time series.

Since the survey is longitudinal, the effects and biases introduced and treated in 2020 continue to be seen in 2021, in particular for the subsamples interviewed in 2020 and re-interviewed in 2021.

Luxembourg

The results apply to those people living in Luxembourg who have been educated in Luxembourg, as well as to those who have been educated in another country. This means the figures cannot be used to analyse the national educational system. There was a break in 2003 due to transition to a quarterly continuous survey (source Eurostat).

New Zealand

Attainment data before 2013 on ISCED levels 4 and 5 are no longer reliable, and trends should not be used. Trend data for total tertiary attainment (ISCED 5 and above), or for the combined ISCED 3 and 4 group, can therefore also no longer be used. Trend data for “below upper secondary”, “upper secondary” only, and “degree and above” can still be reliably used.

Attainment data for New Zealand are sourced from the New Zealand Household Labour Force Survey (HLFS). The educational attainment question in this survey changed in 2013. These changes brought the questions asked more in line with the way educational attainment is collected in other government surveys and administrative collections. The new question provided a more accurate way to map responses to both the New Zealand qualifications framework (NZQF), and to the ISCED. The previous question included a number of categories that related to types of qualifications that could span more than one educational level. For example, “University certificate or diploma below degree level”, “Teacher and nursing certificates or diplomas” or “NZ certificate or diploma”, which can span a number of ISCED levels from 2 to 6. For the reporting of New Zealand attainment data in Education at a Glance, a best-fit mapping of these categories to ISCED was developed, using a method that minimised the level of error inherent form assigning categories spanning more than one level to just one level. The new educational attainment question introduced into the HLFS in 2013 provides a more exact mapping to NZQF levels, and to ISCED levels. In particular, the new question can separate post-initial school ISCED 3, ISCED 4 and ISCED 5 attainment more accurately than the previous question could.

The main impact of this survey change on New Zealand attainment data for Education at a Glance affects ISCED 4 and ISCED 5. Between EAG 2014 (using 2012 data and the old HLFS question) and EAG 2015 interim report (using 2013 data and the new HLFS question), attainment of the 25-64 year-old population at ISCED 5 shifted from 15% to 11%; and at ISCED 4 from and 8% to 16%. Consequently, the proportion with “tertiary” attainment shifted from 40% to 35%. The changes did not significantly affect the proportions with degrees above (ISCED 6 and over), or those with upper secondary (ISCED 3) only, or those with less than upper secondary.

Around 9% of adults in New Zealand have a one-year upper secondary level qualification as their highest attainment. These include the National Certificate of Educational Achievement Level 1, School Certificate for older adults, or a Level 1 National Certificate. Under either ISCED-97 or ISCED 11 these do not count as upper secondary attainment. In earlier editions of Education at a Glance using ISCED-97, these were identified separately as ISCED 3CS. Under ISCED 2011, being used from EAG 2015 on, these one-year qualifications are no longer recognised separately, and are grouped with those with no school qualifications.

Norway

A break in time series on educational attainment occurred in 2005, as the classification of educational attainment was reclassified. Attainment numbers for 2000-2004 follow the former classification of educational attainment and are not comparable with more recent years. The main change is an increase in ISCED 2 attainment, at the expense of ISCED 3. The attainment criteria for ISCED 3 were tightened from course completion to successful completion of the whole programme (studiekompetanse/fagbrev). A reasonable amount of movement also occurred between ISCED 3 and ISCED-97 – level 5, but the net difference is marginal. A minimum of two years full-time study load, equivalent to 120 credit points, is defined as an attainment criterion for ISCED-97 - level 5 (http://www.ssb.no/english/subjects/04/01/utniv_en/).

In 2020, the register on educational attainment of the population was supplemented with new information about immigrants’ highest level of education from the Norwegian Agency for Quality Assurance in Education (NOKUT), the Norwegian Labour and Welfare Administration (NAV) and from the statistical agencies in Iceland, Finland and Denmark. This has caused changes to the proportion of the population by educational attainment. The register will from 2020 be updated annually with data from NOKUT and NAV.

Poland

From 2006 onwards previous 3CS programmes for Poland have been reallocated to 3C Long, since 3C programmes in Poland last three years, which is similar to the typical cumulative duration of a standard national ISCED 3A general programme.

Since the first quarter of 2021, there have been introduced methodological changes in the LFS resulting from the implementation of the framework regulation for social statistics (Regulation (EU) 2019/1700 of the European Parliament and of the Council of 10 October 2019 and its implementing acts).

The changes concern mainly:

• the subjective range – the core part of the survey covers persons aged 15-89 years (until the fourth quarter of 2020, they were persons aged 15 years and more), for other household members, i.e. persons aged below 15 years and over 89 years, there is only collected information regarding the general characteristic of a household,

• definition and the way of specifying particular populations of persons on the labour market – employed, unemployed and economically inactive persons (i.e. in the wording and the order of asking questions),

• the objective range of the survey (partial exchange of the variables).

Therefore, the LFS data since the first quarter of 2021 must not be compared with previous periods.

Sweden

There are the following breaks in the series: 2013/2014 when the ISCED Classification 2011 has been introduced. Upper secondary education with duration shorter than 2 years (mainly AMU-education and Komvux) was reported at ISCED 3 in ISCED-97 and is reported at ISCED 2 according to ISCED 2011. Due to this change, there is a big increase in the share of the population aged 25-64 with “below uppers secondary education”, from 12% to 18%. Furthermore, in 2001 when the new standard for classification of education (SUN 2000) was applied in 2001, and in 2005, when a new EU-harmonised questionnaire was introduced, the break in the series leads, among other consequences, to a breakdown of ISCED-97 levels 4 and 5B into two separate variables. The latter explains the decrease in tertiary attainment 2005.

In 2020, the register on educational attainment of the population was supplemented with new information about immigrants’ highest level of education from the Norwegian Agency for Quality Assurance in Education (NOKUT), the Norwegian Labour and Welfare Administration (NAV) and from the statistical agencies in Iceland, Finland and Denmark. This has caused changes to the proportion of the population by educational attainment. The register will from 2020 be updated annually with data from NOKUT and NAV.

Switzerland

Trend data have been revised from 1997 to 2008 to correct an error in the original data source. Changes in ISCED categories 3CS and 3CL were carried over the time series (1997 to 2008). Before 2001, however, ISCED 3CL only partially reflects the reality. It should not be distinguished from other categories of ISCED 3. In general, before 2001, it is not possible to distinguish between the ISCED categories 1 and 2, as well as to the ISCED categories 3 and 4 or that of ISCED 5A and ISCED-97 – level 6.

Türkiye

The 2007 figures were adjusted according to the new census showing a decrease in total population compared to the projections. For the moment no adjustment/revision are available for the previous years. When the new population projections will be ready, the series will be revised back in time, including 2007 figures. It is not correct to compare 2007 figures with previous years.

United Kingdom

An improved methodology introduced in 2009 led to an increase in measured educational attainment. For 25-64 year-olds the effect was an increase of 3.4 percentage points for those with at least upper secondary level education, and 3.4 percentage points for tertiary level attainment. Women aged 60-64 are included from 2009. The back time series was revised in 2008, taking account of reweighted (to mid-census population estimates) and revised (now using calendar rather than seasonal quarters) data. The revisions provided an opportunity to correct some long-standing anomalies in older data (reported up to 2005), such as an overestimation of the proportion holding ISCED-97 – level 6 (doctoral level), and where ISCED 3B was incorrectly grouped in 3A.

United States

Please note that the NEAC submission for the United States includes those members of the US military services who live in civilian households. These individuals are normally excluded from civilian labour force computations produced by the United States. The figure for inactive population calculated by the OECD as the difference between the total population and those employed in civilian jobs plus unemployed population does include these military service personnel as “inactive”. For this reason, the figure for inactive population differs from the one produced nationally in the United States from the same survey.

Caution should be used when comparing 2022 estimates to those of prior years due to the impact that the coronavirus pandemic had on interviewing and response rates. For additional information about the impact of the coronavirus pandemic on the Current Population Survey data collection, see https://www2.census.gov/programs-surveys/cps/techdocs/cpsmar22.pdf.

Argentina

Data collected through the Permanent Household Survey (Urban) (Encuesta Anual de Hogares Urbanos (EAHU)) covers the 31 Urban Agglomerations as well as all administrative units with 2000 or more inhabitants with the exception of the provinces Tierra del Fuego, Antártida e Islas del Atlántico Sur.

Bulgaria

In 2015, the educational programs corresponding with ISCED codes 342 and 352 did not exist. Since 2021, data on work status of population are not fully comparable with that for the previous years due to changes in the definitions stemmed from the implementation of the new EU regulation for social statistics, in particular in the field of labour force domain, in force from 2021.

Romania

In 2021 LFS was redesigned, thus data for 2021 onwards are not comparable with those for the previous years.

Standard errors

Table X3.A1.3 (see StatLink under Table 1.1) presents estimates and their associated standard errors for data published in Education at a Glance (EAG).

To get a sense of the impact of these standard errors on the meaning and interpretation of the values in the publication it is helpful to compute the associated confidence intervals. These confidence intervals seem reasonably close to the value reported in most cases, indicating that we can be confident about the statistical accuracy of the values in Tables A1.1 and A1.2 and the supplemental educational attainment estimates from OECD.Stat using the available information on sample sizes. However, even though these estimates are relatively precise, small standard errors can still complicate some types of interpretations of these values, in particular, OECD rankings, since small standard errors result in narrow ranges for confidence intervals.

Employing the simple random survey assumption offers a conservative, “best-case scenario” of standard error estimates. As most, if not all, countries’ labour force surveys use complex sample designs, the standard errors would generally be larger if the sample design information were used. The generally small standard errors in Tables A1.1 and A1.2 result in the finding that most of the values are statistically significantly different from the OECD average. If the standard errors were larger, indicating a wider range of possible true values, it would be harder to discern a significant difference between one country and the OECD average value.

While the findings generally support the validity of the tables appearing in EAG, they also suggest that more attention to statistical testing and statistical validity is needed, particularly when detailed data using smaller segments of the population are presented. Also, the standard error estimates should incorporate appropriate adjustments for survey design effects, where the information is available.

For all countries with data available on the sample size, the standard errors were computed under the assumption of a simple random sample. Standard errors taking into account the complex sample design may be higher. The extend of the estimation error of the standard error can be assessed by comparing the precise standard errors released for selected indicators in EAG taking into account the complex survey design (i.e. those provided by a few countries and included in the table) with the respective standard errors computed under the simple random sample assumption in the Education at a Glance database.

Indicator A2. Transition from education to work: Where are today’s youth?

Methodology

Data from the TRANS survey refer to the first quarter of each year (i.e. January, February, March). There are some exceptions for countries reporting the second quarter (i.e. April, May, June), reporting spring quarters (i.e. March, April, May) or the whole year (i.e. January to December). See Table X3.A2.2 1 (see StatLink under Table 1.1) summarising the metadata by country.

“In education” refers to adults in formal education and training.

The calculation of educational attainment for those “in education”, as well as for those “not in education” has changed since the reference year 2006. From this year onwards the ISCED level refers to the completed level of education rather than the attended level of education. People with no information on their educational attainment are excluded from all data disaggregated by educational attainment.

The definition of inactive is that inactive individuals are those who are, during the survey reference week, neither employed nor unemployed, i.e. individuals who are not looking for a job, and therefore there is nothing on whether the person is in education or not. For example, in Table A2.1 column 3, among 18-24 year-olds 18% are in education and employed, therefore those in education should not be considered as active or inactive by default.

Data in Indicators A2 and A3 are not directly comparable due to the difference in reference period (one quarter versus one year). This is particularly important for Indicator A2 which focus on the young adults, given that they might be temporarily out of education during summer period.

Source

The OECD Network on labour market, economic and social outcomes of learning (LSO Network) compiles data on population, educational attainment and labour-market status from national Labour Force Surveys, and usually refer to the first quarter, or the average of the first three months of the calendar year.

The sources for data on the transition from education to work (i.e. Tables A2.1 and A2.2) are the same as in Table A1.1 except for France where the source is the European Union LFS (EU-LFS) for year reference 2009 to 2015 and for the United States where the source is the October Supplement to the Current Population Survey (CPS) instead of the March Supplement. Data for tables A2.3 and A2.4 on recent graduates were extracted from EU-LFS for all countries participating in this survey by Eurostat and the national Labour Force Survey for the United Kingdom.

Table X3.A2.2 1 (see StatLink under Table 1.1) summarises the metadata by country.

Notes on specific countries

Australia

Australian data at the detailed level may be unreliable due to the suppression of small values. The data is indicative only and should be used with caution.

Canada

The Labour Force Survey establishes whether or not a respondent is attending an educational establishment (includes primary, secondary, college, CEGEP, university).

Finland

At the beginning of 2021, the data content, data collection and estimation method of the Finnish Labour Force Survey were revised. The most important content changes concerned working time data and the definition of employed person. The data collection was revised by introducing a web questionnaire as a new data collection method alongside computer-assisted telephone interviews and face-to-face interviews. The revised estimation method utilises more widely different registers, such as Statistics Finland's Register of Completed Education and Degrees.

Revisions of the Finnish Labour Force Survey were not related to the COVID-19 pandemic. The revisions were general development of the Labour Force Survey and due to the revisions of the Integrated European Social Statistics. The statistics produced with the old estimation method are not comparable with the statistics produced with the new estimation method. There is a break in time series in 2021 in the data which are based on the Labour Force Survey.

Germany

The non-response rate was further reduced compared to both previous years.

Increased immigration in 2022 is not fully reflected in the microcensus, which may have effects on the interpretation of results.

The limitations in the survey implementation of the microcensus from 2020 and 2021 were no longer present in the survey year 2022. That is, the initial technical difficulties following the methodological re-design of the microcensus in 2020 and the effects of the Corona crisis did not have any significant effects on the 2022 microcensus. This is reflected, among other things, in the non-response rate, which was further reduced compared to the last two years. The non-response rate for the final results of the 2022 microcensus is approximately 11 % at the federal level, which is already significantly lower than for the final results in 2020 and 2021 (final result in 2021: 14%; final result in 2020: 35%).

In the case of the Microcensus 2022, it must be taken into account that increased immigration in 2022, especially as a result of the Russian attack on Ukraine, may have an impact on the results: In the extrapolation, selected characteristics of the microcensus are adjusted to key figures of the population update, including citizenship. Due to the strong influx of people seeking protection from Ukraine in 2022, the microcensus did not fully record them. In the current population update, however, these groups of people are taken into account via the reports from the registration offices. When interpreting the results on the population without German citizenship, it should therefore be noted that the different foreign citizenships (especially EU third countries) may be over-estimated and that Ukrainian citizenship in particular is underestimated. Moreover, this may also have effects on other characteristics of the microcensus, such as household structure and educational attainment.

Greece

In 2021 the Implementing Regulation (ΕΕ) 2019/2240 for the Labour Force Survey, introduced important changes in the survey methodology. The main changes concern:

• Data collection (general use of computer-aided interviews)

• The formulation of questions related to the employment status of respondents during the reference week (due to the adoption of a model questionnaire prepared by Eurostat)

• The computation of the weighting factors

• The definition of employment and, in particular, the treatment of person reporting having a job during but not working even for one hour.

Israel

Conscripts into the army are considered to be employed, as opposed to 2011 and before, when they were counted as not in the labour force.

Work-study programmes apply to a very small part of the population (currently 4% of secondary students are enrolled in such programmes).

Slovak Republic

Since 2021, the IESS methodology has been applied and a new weighing method was used.

Spain

Those aged 15 are considered in lower secondary level of education and out of labour force because education is compulsory for this age.

The new regulations in the LFS were implemented in 2021 but no significant impact in the Spanish results was detected.

Switzerland

Questionnaires since 2021 was modified to comply with the new Eurostat regulations, also in force from January 2021. For certain results, these changes have led to a break in the series between 2020 and 2021.

United Kingdom

Raw data before 2013 concern 16-29 year-olds. Those aged 15 were previously estimated as the fraction of 1/14 of the total 16-29 year-old population. They are considered in education, with lower secondary level of education and out of labour force.

The work-study programmes definition includes:

• Government employment or training schemes (youth training programme, training for work, action for community employment, job skills, national young traineeship).

• Those on a new deal scheme, working for an employer in the public or private sector, working for the voluntary sector, working for an environmental task force, other type of new deal schemes involving practical training (practical training, at college, temporarily away from project/college).

• Those on the following government employment or training schemes: in England/Wales on a scheme run by a training and enterprise council, in Scotland on a scheme run by a local enterprise company.

• Anyone on a recognised trade apprenticeship not included in any of the above schemes.

The category “Other employed” includes people in education, who are employed but not included in the work-study programme.

In 2022 education variables in the Labour Force Survey were revised which resulted in a decreased sample size for the first quarter of 2022. The revision meant respondents’ previous quarters data for qualification-based questions could not be brought-forward from the fourth quarter of 2021 which typically constitutes 25% of the total.

Standard errors

Table X3.A2.2 (see StatLink under Table 1.1) presents estimates and their associated standard errors for data published in Education at a Glance (EAG).

For more information see above in Indicator A1 the section on standard errors.

Indicator A3. How does educational attainment affect participation in the labour market?

Methodology

The methodology for this Indicator is similar to the information included for Indicator A1 above.

Source

The sources for data in this indicator are the same as for Tables A1.1, A1.2 and A1.3 on educational attainment in Indicator A1.

For further information on sources and notes on specific countries, see Indicator A1.

Standard errors

Tables X3.A3.1 and X3.A3.2 (see StatLink under Table 1.1) present estimates and their associated standard errors for data published in Education at a Glance (EAG).

For more information see above on Indicator A1 the section on standard errors.

Indicator A4. What are the earnings advantages from education?

Methodology

Indicator A4 provides data on earning advantages from education.

Relative earnings of employed compared to employed with upper secondary education are calculated as follows:

$\frac{{\mathit{E}\mathit{a}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{g}\mathit{s}}_{\mathit{A}\mathit{g}\mathit{e},\mathit{G}\mathit{e}\mathit{n}\mathit{d}\mathit{e}\mathit{r},\mathit{E}\mathit{d}\mathit{u}\mathit{c}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}}}{{\mathit{E}\mathit{a}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{g}\mathit{s}}_{\mathit{A}\mathit{g}\mathit{e}.\mathit{G}\mathit{e}\mathit{n}\mathit{d}\mathit{e}\mathit{r}.\mathit{U}\mathit{p}\mathit{p}\mathit{e}\mathit{r}\mathit{S}\mathit{e}\mathit{c}}}\times \mathrm{ }100$

$EarningsAge,Gender,EducLevel$is the mean earnings of persons within an age group and gender with a certain educational attainment level compared to $EarningsAge,Gender,UpperSec$where the mean earnings of persons of the same age group is upper secondary education as highest level of educational attainment.

Women’s earnings relative to men’s earnings are calculated as follows:

$\frac{{\mathit{E}\mathit{a}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{g}\mathit{s}}_{\mathit{A}\mathit{g}\mathit{e},\mathit{E}\mathit{d}\mathit{u}\mathit{c}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l},\mathit{W}\mathit{o}\mathit{m}\mathit{e}\mathit{n}}}{{\mathit{E}\mathit{a}\mathit{r}\mathit{n}\mathit{i}\mathit{n}\mathit{g}\mathit{s}}_{\mathit{A}\mathit{g}\mathit{e}.\mathit{E}\mathit{d}\mathit{u}\mathit{c}\mathit{L}\mathit{e}\mathit{v}\mathit{e}\mathit{l}.\mathit{M}\mathit{e}\mathit{n}}}\times \mathrm{ }100$

$EarningsAge,EducLevel,Women$is the annual mean earnings of women of a particular age group and highest educational attainment level and $EarningsAge,EducLevel,Men$is the annual mean earnings of men of the same age group and level of educational attainment.

Level of earnings relative to median earnings:

The level of earnings relative to median earnings is defined as the ratio of the number of people with earnings within an earnings level relative to the median and all persons with earnings from employment. The distribution is calculated for all earners (including full-time, full-year earners and part-time earners). These relative earnings are broken down into the following earning levels:

• At or below half the median

• More than half the median but at or below the median

• More than the median but at or below 1.5 times the median

• More than 1.5 times the median but at or below twice the median

• More than twice the median

Source

The OECD Network on labour market, economic and social outcomes of learning (LSO Network) compiles data on education and earnings.

Tables X3.A4.1 and X3.A4.2 (see StatLink under Table 1.1) summarise the metadata by country.

Table X3.A4.3 (see StatLink under Table 1.1) summarise the information on the treatment of persons with zero or negative earnings.

Table X3.A4.6 (see StatLink under Table 1.1) summarises the coverage of individuals with no earnings from work.

Notes on specific countries

Australia

The data source used since reference year 2016 is different from the one used in previous years. As a result, data are not directly comparable between 2016 onwards and previous years.

Belgium

. For the level of earnings relative to median earnings, the data source is the EU Labour Force Survey (EU-LFS). From 2021 onwards, gross pay is reported and this information is collected from administrative data. Prior to 2021, self-reported earnings net of income tax were used. This change in reporting means that there is a break in the data source between 2021 and the previous years

Canada

The data source used since reference year 2014 is different from the one used in previous years. As a result, data are not directly comparable between 2014 onwards and previous years. For data collected from the Canadian Income Survey (CIS), population estimates of a value equal to or less than 5,000 should not be released, nor used in the calculation of ratios. Characteristics of such populations should not be released either.

Chile

Data on earnings is obtained from a process of inclusion of compulsory salary deductions, depending on labour and social characteristics of each observation.

Czech Republic

The term full-time is a self-designated full-time status. Working hours are defined for a concrete position which is the same as real time usage defined as a full-time. As far as the working hours defined for concrete job differ from real time the employee spends at work, it is defined as part-time. There is another additional criterion that says: if the defined working hours for concrete position are less than 30 hours per week, it automatically marked as a part-time. But the usual working time is 40 hours per week for full-time.

Finland

Data on earnings by field of study for 2017 and 2020 used a different data source. In 2017 earnings data by fields was from the Structure of Earnings Survey while the latest dataset is from the Employment Statistics.

France

Since reference year 2012, the age is measured at the beginning of the reference period, i.e. one year before the survey. Age was previously measured at the beginning of the survey year.

Ireland

Regarding data for year 2020 (data collection in 2021), there has been a complete re-design of SILC (Survey on Income and Living Conditions) questionnaire for Ireland and processes resulting in breaks in series in certain variables.

Latvia

Starting from 2021 wages and salaries is an annual indicator, i.e., are only collected at the first-wave interview and the estimates are based on the first-wave weights. Starting from 2021 gross wages and salaries, until 2020 – net wages and salaries. Information on wages not specified as well as exact amount of wages indicated within the given pay ranges was drawn from the State Revenue Service database.

Luxembourg

The data source used since reference year 2016 is different from the one used in previous years. As a result, data are not directly comparable between 2016 onwards and previous years.

Norway

Information on those working full-time, full-year is collected from an administrative register on employees; the Employer and Employee register (EE-register). For reference years before 2014, full-time full-year earners are defined as those being registered with a job each month through the year with a contractual number of at least 35 hours per week each month. Since reference year 2015, full-time full-year earners are defined as those being registered with a job in December with a contractual number of at least 35 hours per week.

The EE-register covers about 90 per cent of all the employees. Those not covered are mainly employees with short term jobs.

The EE-register has not been used for compiling these kinds of data so far. There are some quality problems with the EE‑register which probably results in an overestimation of the number of full-time, full-year employees. Updating of the EE‑register is done by the employers. Some employers might have forgotten to report about employees that have left their job before the end of the year and some might have forgotten to report about employees that have decreased their contractual hours below 35 hours.

Portugal

The data, whose source is "Personnel Tables", refers to firms with employees in the private sector (covered by Portuguese Labour Code), excluding public administration and the coverage of the agricultural sector is low. Civil servants (with employment contracts in public functions), self-employed and domestic service are not covered. Data refers to employees (and respective earnings in the main job) at establishments in Portugal in October 2020.

Sweden

Data from total registers for year 2017 while previous years data was from EU Statistics on Income and Living Conditions (EU-SILC), therefore there is a break in time series. The distribution of full-time full-year and part-time part-year earners have changed due to a methodological change. A new source “AGI Employer declaration at individual level” is used, which has much higher quality than the previous used “KU register”. In the old source, the number of full-year workers was over-estimated. The effect on the earnings data is a decline in full-time full-year workers and a corresponding increase of part-time part-year workers, as a share of all workers. The Indicator A4 on full-time earners might be slightly affected but measures based on all earners or total population should not be affected at all. The category "No earners" has been redefined from year 2020. Before 2020 people with zero earnings has been classified as No earners no matter if they worked or not. From 2020 the number of months worked is decisive, rather than the level of earnings. This resulted in 52 000 people transferred from "No earnings" to "All earners". About 10% of them are full-time full-year earners. The overall impact is expected to be low. For example, the effect on median earning is -0,5 percent.

United Kingdom

Data for upper secondary attainment include completion of a sufficient volume and standard of programmes that would be classified individually as completion of intermediate upper secondary programmes (about 15% of the adults are in this group).

United States

Data users should exercise caution when comparing Current Population Survey, Annual Social and Economic Supplement estimates for data years 2019, 2020, and 2021 from the reports or from the microdata files to those from previous years due to the effects that the coronavirus (COVID-19) had on interviewing and response rates. Using administrative data, U.S. Census Bureau researchers have documented that nonrespondents to the Current Population Survey in 2020 to 2022 are less similar to respondents than in earlier years. Among these differences, of particular interest is that respondents from 2020 to 2022 had relatively higher income than non-respondents. Actual earnings data

Tables X3.A4.4 and X3.A4.5 (see StatLink under Table 1.1) present earnings data in actual values in USD using Purchasing Power Parity (PPP) for private consumption.

Indicator A5. What are the incentives to invest in education?

This indicator was not included in Education at a Glance (EAG) 2023.

Indicator A6. How are social outcomes related to education?

Methodology

Table A6.1. Average score for the perception of democracy, by educational attainment (2020)

Additional data on national elections being free and fair, and media freedom are available for consultation online (https://stat.link/szkel2). Additional data on views on governing parties being punished in elections, government measures to reduce differences in income levels, and the will of the people can also be found online (https://stat.link/szkel2).

Data from the European Social Survey were extracted from Round 10, published in June 2022. More information on the methodology of this survey is available at: https://www.europeansocialsurvey.org/methodology/

For this table, respondents answered the following question:

 

In Table A6.1, the options of the protection of the rights of minority groups and political parties offering clear alternatives to one another were not considered to assess perception of democracy. Other options that do not appear among the statements above but that are included in the Question CARD 36 are: the will of the people cannot be stopped, governing parties are punished in elections when they have done a bad job, the government takes measures to reduce differences in income levels.

Table A6.2. Share of adults who reported the following behaviour indicating civic engagement, by educational attainment and programme orientation (2020)

Additional data on the share of adults who reported having taken part in a public demonstration and those reported having volunteered for a not-for-profit or charitable organisation can be found online (https://stat.link/hqyrxa).

For this table, respondents answered the following question:

 

Data from the European Social Survey were extracted from Search European Social Survey (nsd.no). More information on the methodology of this survey is available at: https://www.europeansocialsurvey.org/methodology.

Table A6.3. Share of Internet users taking precautions to protect the privacy of their personal data, by educational attainment and type of precaution (2021)

Additional data on limiting access and use of data, checking whether websites were secure or asking administrators to update personal data are available for consultation on line (https://stat.link/jkbglc).

For this table, respondents answered the following question:

 

Data on the EU Survey on ICT usage in households and by individuals (EU-ICT) survey were extracted from Database - Digital economy and society - Eurostat (europa.eu). More information on the methodology of this survey is available at: https://ec.europa.eu/eurostat/cache/metadata/en/isoc_i_esms.htm

Table A6.4 (Web only). Average score for the perception of democracy, by educational attainment and gender (2020)

Data from the European Social Survey were extracted from Search European Social Survey (nsd.no). More information on the methodology of this survey is available at: https://www.europeansocialsurvey.org/methodology.

Table A6.5 (Web only). Average score for the perception of democracy, by educational attainment and age group (2020)

Data from the European Social Survey were extracted from Search European Social Survey (nsd.no). More information on the methodology of this survey is available at: https://www.europeansocialsurvey.org/methodology.

Table A6.6 (Web only). Share of Internet users taking precautions to protect the privacy of their personal data, by gender, educational attainment and type of precaution (2021)

Data on the EU-ICT survey were extracted from Database - Digital economy and society - Eurostat (europa.eu). More information on the methodology of this survey is available at: https://ec.europa.eu/eurostat/cache/metadata/en/isoc_i_esms.htm

Table A6.7 (Web only). Share of Internet users taking precautions to protect the privacy of their personal data, by type of precaution, age group, and educational attainment (2021)

Data on the EU-ICT survey were extracted from Database - Digital economy and society - Eurostat (europa.eu). More information on the methodology of this survey is available at: https://ec.europa.eu/eurostat/cache/metadata/en/isoc_i_esms.htm

Table A6.8 (Web only). Share of adults reporting they agree or strongly agree with the following statements, indicating belief in conspiracy theories (2020)

Variables included in this table are based on the European Social Survey Round 10 for European OECD member countries.

Data from the European Social Survey were extracted from Search European Social Survey (nsd.no). More information on the methodology of this survey is available at: https://www.europeansocialsurvey.org/methodology.

For this table, respondents answered the following question:

 

 

Notes on specific countries

Canada

• Table A6.2 : Data for the share of adults (25-64 year-olds) who reported having : taken part in a public demonstration, boycotted certain products, posted or shared anything about politics on line, during the last 12 months, is from 2020. Data for the share of adults that volunteered for a not-for-profit or charitable organisation, is from 2018.

Source

Table A6.1. Average score for the perception of democracy, by educational attainment (2020)

• European Social Survey (ESS), Round 10

Table A6.2. Share of adults who reported the following behaviour indicating civic engagement, by educational attainment and programme orientation (2020)

• European Social Survey (ESS), Round 10 for OECD and accession countries participating in this survey

• The General Social Survey (GSS SI and GSS GVP) for Canada

Table A6.3. Share of Internet users taking precautions to protect the privacy of their personal data, by educational attainment and type of precaution (2021)

• EU Survey on ICT usage in households and by individuals (EU-ICT) for OECD and accession countries participating in this survey

• The Canadian Internet Use Survey (CIUS) 2020 for Canada

Table A6.4 (Web only). Average score for the perception of democracy, by gender, educational attainment and programme orientation (2020)

• European Social Survey, Round 10 for European OECD member countries and Israel

Table A6.5 (Web only). Average score for the perception of democracy, by educational attainment and age group (2020)

• European Social Survey, Round 10

Table A6.6 (Web only). Share of Internet users taking precautions to protect their privacy of personal data, by gender, educational attainment and type of precaution (2021)

• EU Survey on ICT usage in households and by individuals (EU-ICT) for OECD and accession countries participating in this survey

• The Canadian Internet Use Survey (CIUS) 2020 for Canada

Table A6.7 (Web only). Share of Internet users taking precautions to protect the privacy of their personal data, by age group, educational attainment and type of precaution (2021)

• EU Survey on ICT usage in households and by individuals (EU-ICT) for OECD and accession countries participating in this survey

• The Canadian Internet Use Survey (CIUS) 2020 for Canada

Table A6.8 (Web only). Share of adults reporting they agree or strongly agree with the following statements, indicating belief in conspiracy theories (2020)

• European Social Survey (ESS), Round 10

Indicator A7. To what extent do adults participate in education and training?

Methodology

This indicator includes data on participation in formal and/or non-formal education and training from different sources that have different reference period: either 4 weeks or 12 months before the survey.

The European Union Labour Force Survey (EU-LFS) is held quarterly and measures participation in formal and/or non-formal education and training during a four-week period excluding guided on-the-job training. The EU-LFS methodology can be found at https://ec.europa.eu/eurostat/statistics-explained/index.php?title=EU_labour_force_survey_-_methodology. The reference period for participation in non-formal education and training is during the previous 4 weeks for the EU-LFS and the United Kingdom national surveys and the previous 12 months for the Australia and Costa Rica national surveys.

The European Union Continuous Vocational Training Survey (EU-CVTS) takes place every five year and measure the on continuing vocational training carried out in enterprises over the 12 months prior to the survey. The EU-CVTS methodology can be found at: https://ec.europa.eu/eurostat/statistics-explained/index.php?title=Continuing_Vocational_Training_Survey_(CVTS)_methodology. National survey of Switzerland measures training costs in enterprises during a 12-months period.

Source

Table A7.1 Share of adults participating in non-formal education and training, by labour-market status, job-relatedness and gender (2022)

• EU Labour Force Survey (EU-LFS) for European OECD and accession countries (Austria, Belgium, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden, Switzerland and Türkiye).

• National surveys for Australia (Australian Bureau Survey of Work-Related Training and Adult Learning), Costa Rica (Continuous Employment Survey), Canada (Labour Force Survey), Korea (Korean Adult Lifelong Learning Survey) and the United Kingdom (Labour Force Survey).

The reference period for participation in non-formal education and training is during the previous 4 weeks for countries using the EU-LFS and the previous 12 months for the Australia, Canada, Costa Rica and Korea national surveys.

Table A7.2 Share of adults participating in non-formal job-related education and training, by age group, educational attainment and programme orientation (2022)

• EU Labour Force Survey (EU-LFS) for European OECD and accession countries (Austria, Belgium, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden and Switzerland).

• National surveys for Australia (Australian Bureau Survey of Work-Related Training and Adult Learning), Canada (Labour Force Survey) and Korea (Korean Adult Lifelong Learning Survey).

The reference period for participation in non-formal job-related education and training is during the previous 4 weeks for countries using data from the EU-LFS and the previous 12 months for Australia, Canada and Korea.

Table A7.3 Training costs as a share of total labour costs, by size of enterprise (2010, 2015 and 2020)

• The Continuing Vocational Training Survey (CVTS) for European OECD and accession countries (Austria, Belgium, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden) and the United Kingdom.

• National survey for Switzerland (Swiss Continuing Education and Training Survey).

  The reference period for training costs is the previous 12 months for all countries with available data.

Table A7.4 Share of employed adults participating in non-formal job-related education and training, by size and sector of enterprise (2022)

• EU Labour Force Survey (EU-LFS) for European OECD and accession countries (Austria, Belgium, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden and Switzerland).

• National survey for Canada (Labour Force Survey).

  The reference period for participation in non-formal job-related education and training is during the previous 4 weeks for countries using data from the EU-LFS and the previous 12 months for Canada.

  Table A7.5 Share of employed adults participating in non-formal job-related education and training, by economic activity (2022)

• EU Labour Force Survey (EU-LFS) for European OECD and accession countries (Austria, Belgium, Bulgaria, Croatia, the Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Iceland, Ireland, Italy, Latvia, Lithuania, Luxembourg, the Netherlands, Norway, Poland, Portugal, Romania, the Slovak Republic, Slovenia, Spain, Sweden and Switzerland).

• National survey for Canada (Labour Force Survey).

  The reference period for participation in non-formal job-related education and training is during the previous 4 weeks for countries using data from the EU-LFS and the previous 12 months for Canada.

Notes on specific countries

Canada

In Table A7.4 on the share of employed adults participating in non-formal job-related education and training, by size and sector of enterprise, the size of enterprise differs as follows: 10-49 employed persons includes 20-99 employed persons; 50-249 employed persons includes 100-500 employed persons; over 249 employed persons includes over 500 employed persons.

References

[1] OECD (2018), OECD Handbook for Internationally Comparative Education Statistics 2018: Concepts, Standards, Definitions and Classifications, OECD Publishing, Paris, https://doi.org/10.1787/9789264304444-en.