Parallel session C
How history shapes adult skills: Studies on cohorts, educational reforms and survey quality
Session room: RUUSUPUISTO LOBBY
The long shadow of education policy illustrated by historizing PIAAC results – an exploratory case analysis
lorenz lassnigg, IHS Institute for Advanced Studies, Vienna, Austria
Stefan Vogtenhuber, IHS Institute for Advanced Studies, Vienna, Austria
Historising PIAAC means to analyse the production machinery that has brought about the measured cross-sectional skills in the population over the decades since the oldest cohorts have started school in the 1950s. This illustrates the dimension of this big machinery that includes the formal education careers plus all subsequent learning experiences. Thus, PIAAC indicates the results not of current practices but of very different structures and policies over time. The aim of the study is a stylized exploratory in-depth analysis of traces of past educational structures and policies in the age-specific profile of PIAAC results. The background for this national in-depth analysis is a previous comparative analysis of the first round of PIAAC that traced measured skills to the two educational reform waves, expansive-egalitarian in the 1960s-70s, and neoliberal in the 1980s-90s. Austria was compared to selected countries from different welfaregimes (Nordic, Liberal, Continental). Focus was on understanding the Austrian model of skill formation in comparative terms. Quite diverse results for countries were gathered. Now, the second wave of PIAAC is used to look more detailed into the Austrian case (might be exemplary for other cases). The stylised framework fits into a broad theoretical landscape including the political economy of skill formation, redistributive and cultural political economy, historical institutionalism, and complexity theory. The methods use a national case analysis because the specific education structures must be identified. Part 1 provides a multifactorial understanding and reconstruction of the development of educational structures and policies, based on selected stylised factors: reform-dynamic, resources, demography, migration, expanding participation, assessment. Part 2 confronts this historical model with the PIAAC results from the first and second wave, using indicators of aggregate 5-years age-groups (scores for skills level and percentile-rates for skills distribution). PIAAC indicators are analysed by the age groups at the two waves, and by a quasi-cohort comparison that follows some age groups up to the second wave. A main result and a tool for interpretation is a detailed time-specific model of the main educational changes over the affected decades since the 1950s, related to long-term skills outcomes (of course belated with all changes and effects after formal education). A basic result is the high reform dynamic, so the PIAAC population was educated over time in five different education structures. Marked long term effects of these reforms are not found. Austerity phases do not show marked declines of PIAAC results, and expansion of upper-level education does neither show dampening skill levels nor increase of inequality. A small tendency of neoliberal policy shift indicates decline of skills rather than improvement. The analysis is preliminary and shows what can be learned from historizing PIAAC.
Age-group comparisons and competence profiles in PIAAC 2
Panu Forsman, University of Jyväskylä, Finland
Maria Fisk, University of Jyväskylä, Department of Education, Finland
Working-age competencies are societal concern reshaped by technological change, globalization, and volatile markets (Marillo et al., 2024). Knowledge-intensive jobs require strong cognitive and personal skills (Giddens, 2007), while weakening social structures and guidelines contribute to uncertainty and reflexive imperative (Bauman, 2007; Archer, 2012). Workforce consists of generations with differences in education and socialization background. Although assessments such as PIAAC enable cross-national comparisons, generational differences as historically constituted patterns have received less in-depth research. Nordic countries offer a useful comparison. Despite shared welfare-state roots, Finland, Sweden, and Denmark differ in demographic, economic, education and adult-learning arrangements. Using PIAAC 2 country datasets, we operationalize “pseudo-generations” via OECD 10-year age bracket (AGEG10LFS) found in Finnish, Swedish and Danish sets. In addition, cluster analysis is used to identify subgroups of adults in the named datasets utilizing plausible values for literacy, numeracy and adaptive problem solving variables. Generated competence profiles are then compared against background variables to find potential differences in generational and national levels. The goal is to form historically grounded interpretations of how education systems and working life have shaped competence across cohorts. Age is analyzed using the OECD 10-year age-group classification, for which, the approximate birth-year ranges are derived to connect cohort patterns to period-specific contexts of schooling and work.
Illustrative preliminary generational cohort framings:
• 16–24: digitalized schooling; distance learning; high-speed networks/AI (Gen Y / iGen).
• 25–34: expanded higher education; mobile technologies (Millennials).
• 35–44: educated during restructuring; adulthood in a digital world (Xennials/Millennials).
• 45–54: industrial–post-industrial transition; early internet-era working life (Gen X).
• 55–65: schooling largely pre-digitalization (Baby Boomers).
Analyses are conducted in R. Proficiency comparisons use OECD plausible values (e.g., PVLIT1–PVLIT10 for literacy and corresponding numeracy/adaptive problem-solving PVs). Research questions are:
1. How do proficiency levels differ by cohort in Finland, Sweden, and Denmark?
2. What cohort-specific competence profiles emerge?
3. Which background factors help explain the differences?
As preliminary findings, K-means clustering on PVAPS1–10 identified three groups with clearly separated proficiency levels high, medium and low. Age distributions differed significantly across proficiency profiles, with a small-to-moderate association. Group comparisons indicated that younger adults (≤44) were over-represented in the high-APS cluster, whereas the 55+ group was markedly over-represented in the mid-APS cluster and strongly under-represented in the high-APS cluster.
Did Illiteracy Double in Poland? Response Processes and Interviewer Effects in PIAAC 2023
Michal Sitek, Educational Research Institute - Public Research Institute (IBE PIB), Poland
Katarzyna Chyl-Tanas, Educational Research Institute - Public Research Institute (IBE PIB), Poland
PIAAC Cycle 2 results show a sharp decline in functional literacy in some countries. In Poland, the share of low-skilled adults appears to have doubled. This raises a basic question: does this reflect a real change in skills, or is it partly shaped by how the assessment was administered? The two are not easily separable in low-stakes, home-based settings. Observed performance reflects not only underlying proficiency, but also how respondents engage with tasks and how the session is conducted. Consequently, part of the recorded performance reflects the measurement process itself. We use Polish PIAAC 2023 national data and combine item-level process indicators with operational fieldwork metadata not available in public files, including interviewer identifiers and detailed timing. Because standalone process indicators cannot establish whether careless or insufficient effort responding (C/IER) is structured by fieldwork conditions, we model clusters of response signals. These include compressed response times, item omissions, and atypical response patterns. We estimate mixed-effects models to capture interviewer-level variation and finite mixture models to identify latent response profiles. We distinguish between the interviewer-paced Background Questionnaire (BQ) and the self-administered Direct Assessment (DA) to locate where response quality deteriorates and how interviewer effects extend beyond direct interaction. This distinction turns out to be consequential for interpreting the observed results. Despite extensive quality assurance measures implemented nationally and internationally, the results show clear and systematic variation in indicators of survey quality in the Polish data. Behaviour consistent with C/IER is frequent and structured rather than random. This behaviour varies across interviewers and clusters around specific administration patterns in both BQ and DA components. This indicates that problematic responding is not only a function of individual motivation or ability. It is partly shaped by the fieldwork process itself. When response-process indicators and administration features are modelled jointly, measurement-related distortions are substantial. Cross-cycle differences in PIAAC therefore cannot be interpreted solely as changes in population skills. They also reflect variation in how performance is produced and recorded. Linking process indicators with operational survey data is therefore not optional for interpretation; it changes the substantive conclusion.
The long shadow of education policy illustrated by historizing PIAAC results – an exploratory case analysis
lorenz lassnigg, IHS Institute for Advanced Studies, Vienna, Austria
Stefan Vogtenhuber, IHS Institute for Advanced Studies, Vienna, Austria
Historising PIAAC means to analyse the production machinery that has brought about the measured cross-sectional skills in the population over the decades since the oldest cohorts have started school in the 1950s. This illustrates the dimension of this big machinery that includes the formal education careers plus all subsequent learning experiences. Thus, PIAAC indicates the results not of current practices but of very different structures and policies over time. The aim of the study is a stylized exploratory in-depth analysis of traces of past educational structures and policies in the age-specific profile of PIAAC results. The background for this national in-depth analysis is a previous comparative analysis of the first round of PIAAC that traced measured skills to the two educational reform waves, expansive-egalitarian in the 1960s-70s, and neoliberal in the 1980s-90s. Austria was compared to selected countries from different welfaregimes (Nordic, Liberal, Continental). Focus was on understanding the Austrian model of skill formation in comparative terms. Quite diverse results for countries were gathered. Now, the second wave of PIAAC is used to look more detailed into the Austrian case (might be exemplary for other cases). The stylised framework fits into a broad theoretical landscape including the political economy of skill formation, redistributive and cultural political economy, historical institutionalism, and complexity theory. The methods use a national case analysis because the specific education structures must be identified. Part 1 provides a multifactorial understanding and reconstruction of the development of educational structures and policies, based on selected stylised factors: reform-dynamic, resources, demography, migration, expanding participation, assessment. Part 2 confronts this historical model with the PIAAC results from the first and second wave, using indicators of aggregate 5-years age-groups (scores for skills level and percentile-rates for skills distribution). PIAAC indicators are analysed by the age groups at the two waves, and by a quasi-cohort comparison that follows some age groups up to the second wave. A main result and a tool for interpretation is a detailed time-specific model of the main educational changes over the affected decades since the 1950s, related to long-term skills outcomes (of course belated with all changes and effects after formal education). A basic result is the high reform dynamic, so the PIAAC population was educated over time in five different education structures. Marked long term effects of these reforms are not found. Austerity phases do not show marked declines of PIAAC results, and expansion of upper-level education does neither show dampening skill levels nor increase of inequality. A small tendency of neoliberal policy shift indicates decline of skills rather than improvement. The analysis is preliminary and shows what can be learned from historizing PIAAC.
Age-group comparisons and competence profiles in PIAAC 2
Panu Forsman, University of Jyväskylä, Finland
Maria Fisk, University of Jyväskylä, Department of Education, Finland
Working-age competencies are societal concern reshaped by technological change, globalization, and volatile markets (Marillo et al., 2024). Knowledge-intensive jobs require strong cognitive and personal skills (Giddens, 2007), while weakening social structures and guidelines contribute to uncertainty and reflexive imperative (Bauman, 2007; Archer, 2012). Workforce consists of generations with differences in education and socialization background. Although assessments such as PIAAC enable cross-national comparisons, generational differences as historically constituted patterns have received less in-depth research. Nordic countries offer a useful comparison. Despite shared welfare-state roots, Finland, Sweden, and Denmark differ in demographic, economic, education and adult-learning arrangements. Using PIAAC 2 country datasets, we operationalize “pseudo-generations” via OECD 10-year age bracket (AGEG10LFS) found in Finnish, Swedish and Danish sets. In addition, cluster analysis is used to identify subgroups of adults in the named datasets utilizing plausible values for literacy, numeracy and adaptive problem solving variables. Generated competence profiles are then compared against background variables to find potential differences in generational and national levels. The goal is to form historically grounded interpretations of how education systems and working life have shaped competence across cohorts. Age is analyzed using the OECD 10-year age-group classification, for which, the approximate birth-year ranges are derived to connect cohort patterns to period-specific contexts of schooling and work.
Illustrative preliminary generational cohort framings:
• 16–24: digitalized schooling; distance learning; high-speed networks/AI (Gen Y / iGen).
• 25–34: expanded higher education; mobile technologies (Millennials).
• 35–44: educated during restructuring; adulthood in a digital world (Xennials/Millennials).
• 45–54: industrial–post-industrial transition; early internet-era working life (Gen X).
• 55–65: schooling largely pre-digitalization (Baby Boomers).
Analyses are conducted in R. Proficiency comparisons use OECD plausible values (e.g., PVLIT1–PVLIT10 for literacy and corresponding numeracy/adaptive problem-solving PVs). Research questions are:
1. How do proficiency levels differ by cohort in Finland, Sweden, and Denmark?
2. What cohort-specific competence profiles emerge?
3. Which background factors help explain the differences?
As preliminary findings, K-means clustering on PVAPS1–10 identified three groups with clearly separated proficiency levels high, medium and low. Age distributions differed significantly across proficiency profiles, with a small-to-moderate association. Group comparisons indicated that younger adults (≤44) were over-represented in the high-APS cluster, whereas the 55+ group was markedly over-represented in the mid-APS cluster and strongly under-represented in the high-APS cluster.
Did Illiteracy Double in Poland? Response Processes and Interviewer Effects in PIAAC 2023
Michal Sitek, Educational Research Institute - Public Research Institute (IBE PIB), Poland
Katarzyna Chyl-Tanas, Educational Research Institute - Public Research Institute (IBE PIB), Poland
PIAAC Cycle 2 results show a sharp decline in functional literacy in some countries. In Poland, the share of low-skilled adults appears to have doubled. This raises a basic question: does this reflect a real change in skills, or is it partly shaped by how the assessment was administered? The two are not easily separable in low-stakes, home-based settings. Observed performance reflects not only underlying proficiency, but also how respondents engage with tasks and how the session is conducted. Consequently, part of the recorded performance reflects the measurement process itself. We use Polish PIAAC 2023 national data and combine item-level process indicators with operational fieldwork metadata not available in public files, including interviewer identifiers and detailed timing. Because standalone process indicators cannot establish whether careless or insufficient effort responding (C/IER) is structured by fieldwork conditions, we model clusters of response signals. These include compressed response times, item omissions, and atypical response patterns. We estimate mixed-effects models to capture interviewer-level variation and finite mixture models to identify latent response profiles. We distinguish between the interviewer-paced Background Questionnaire (BQ) and the self-administered Direct Assessment (DA) to locate where response quality deteriorates and how interviewer effects extend beyond direct interaction. This distinction turns out to be consequential for interpreting the observed results. Despite extensive quality assurance measures implemented nationally and internationally, the results show clear and systematic variation in indicators of survey quality in the Polish data. Behaviour consistent with C/IER is frequent and structured rather than random. This behaviour varies across interviewers and clusters around specific administration patterns in both BQ and DA components. This indicates that problematic responding is not only a function of individual motivation or ability. It is partly shaped by the fieldwork process itself. When response-process indicators and administration features are modelled jointly, measurement-related distortions are substantial. Cross-cycle differences in PIAAC therefore cannot be interpreted solely as changes in population skills. They also reflect variation in how performance is produced and recorded. Linking process indicators with operational survey data is therefore not optional for interpretation; it changes the substantive conclusion.