Stability and change in adults' literacy and numeracy skills: Evidence from two large-scale panel studies (original) (raw)

Introduction

In today's knowledge-based and technologically advanced societies, literacy (i.e., the ability to understand, use, and interpret written text) and numeracy (i.e., the ability to access, use, and interpret mathematical information) are quintessential skills1 for the welfare and well-being of individuals and societies. Literacy and numeracy are linked to individual-level outcomes such as income, health, and social participation as well as and macro-level outcomes such as economic growth (Hanushek et al., 2015; Hanushek & Woessmann, 2015; OECD, 2016, e.g., OECD, 2013a).

Demographic aging and technological advances in industrialized societies imply that adults will be increasingly required to update their skills throughout the life course. Literacy and numeracy are key prerequisites to acquiring many forms of knowledge and skills (e.g., reading a machine's manual, programming a computer, or learning a foreign language), and hence for lifelong learning at large. Because not all individuals attain sufficient levels of literacy during schooling age (Durda et al., 2020; Wicht et al., 2021; Wolf & Jenkins, 2014), a key question is whether literacy and numeracy skills can still change during adulthood; and if so, whether change would involve only losses_—_or proficiency gains at least in some adults? Moreover, how is skill change distributed in the population and in socio-demographic subgroups? Potential losses in skills over time in at least some adults might explain why many adult surveys such as PIAAC or LEO find substantial shares of adults with low competences despite these adults having completed 9 or more years of schooling or vocational qualifications (Durda et al., 2020; Grotlüschen et al., 2016, Grotlüschen et al., 2020).

These questions are not only interesting research questions that are relevant to theories of cognitive ageing and intelligence. They are also relevant to policymakers and practitioners interested in promoting lifelong learning (e.g., Wolf & Jenkins, 2014). For example, if literacy and numeracy were impervious to change during adulthood, basic skills or workplace learning programs aimed at fostering literacy and numeracy might have a low return on investment. Childhood might then be a more promising life stage for interventions (e.g., Cunha & Heckman, 2007). Moreover, literacy and numeracy may change over time, but gains and losses may be unevenly distributed across (groups of) individuals. Such findings may help identify segments of the population who are at a heightened risk of experiencing skill loss (e.g., older adults or the lower-educated). These segments may need support from policymakers and practitioners, such as by providing them access to resources and skill programs. Thus, a better understanding of stability and change of literacy and numeracy skills in adulthood may aid the development of targeted policies and interventions in the future.

Given the policy relevance of literacy and numeracy development during adulthood, what answers can extant research provide us?

Prominent theoretical perspectives on skill development in adulthood agree that skills are characterized by lifelong plasticity. However, they differ in their predictions as to when skill change may occur and for whom and whether it involves gains or losses.

In his seminal work, Cattell (1971) hypothesized that “fluid intelligence” (Gf; the ability to process novel information) and “crystallized intelligence” (Gc; the totality of acquired knowledge and skills) follow different age trajectories. He posited that _Gf—_largely innate and dependent on biological functioning—starts to decline after the second life decade. In turn, _Gc—_which results from the investment of Gf in different subject areas and is influenced by education, experience, and culture—continues to increase across adulthood and declines only in very old age. Later research on cognitive aging confirmed and expanded his ideas. For example, Baltes (1993) distinguished between cognitive mechanics (comparable to Gf) and cognitive pragmatics (comparable to Gc). He, too, assumed that cognitive pragmatics reflect the impact of culture and learning and would remain stable until old age. Literacy and numeracy as measured in our present study are conceptualized as skills that are acquired through education and practice in different contexts. As acquired skills, they resemble the conception of Gc or cognitive pragmatics in Cattell's and Baltes's work. The definitions of literacy and numeracy in these surveys also overlaps with reading and writing ability (Grw)2 and quantitative knowledge (Gq), respectively, in the updated Cattell-Horn-Carroll (CHC) model of intelligence, a prominent models of the structure of intelligence (McGrew, 2009). The CHC model, too, classifies Grw and Gq as largely acquired abilities. Hence, based on these theories, it can be expected that literacy and numeracy skills remain largely stable or even increase across adulthood—until perhaps very old age, when physical aging and limitations lead to declines even in acquired skills.

Another perspective from research on adult learning is practice engagement theory [PET; Reder, 1994]. PET originates from literacy research but applies equally to numeracy (Reder et al., 2020). It posits that individuals' literacy skills develop as a by-product of their engagement in everyday reading and writing practices. In turn, it literacy skills affect levels of engagement in reading and writing practices. PET would predict that adults can experience both gains and losses in literacy and numeracy skills over time_—_depending on the extent to which they engage in attendant practices in their job, leisure, or other contexts. This central tenet of PET is compatible with Cattell's and Baltes's perspectives but extends them by highlighting individual differences in skill development and identifying a core drive of these differences.

A third perspective hails from life course sociology: the Matthew effect hypothesis, also known as cumulative (dis-)advantage (Dannefer, 1987; see also (Blossfeld and von Maurice, 2011). Applied to skill development, this perspective would predict that initial skill differences become magnified over the life course. Through the joint effects of selection and socialization, individuals with higher initial skill levels will continue to gain in skills compared to those with initially lower skills. For example, those with higher skills as children will, on average, attain higher educational qualifications than those with lower skills, which in turn channels them into more complex jobs that foster the further development of skills. Like PET, the Matthew effect hypothesis champions individual differences in skills and the relevance of social context in governing skill development.

Together, these theories highlight the lifelong plasticity of (acquired) skills and age-dependency of skill development. Moreover, they emphasize that both gains and losses in skills can occur and highlight individual differences in skill development.

Existing evidence on age differences in literacy and numeracy generally support the theoretical perspectives outlined above (for reviews, see Desjardins & Warnke, 2012; Paccagnella, 2016; also see Deary, 2014), demonstrating, in particular, (1) the lifelong plasticity of skills, (2) life-stage dependency of change, and (3) individual and subgroup differences in skill change.

First, literacy and numeracy appear to continue to develop across adulthood. Despite the fact that literacy (and likely also numeracy skills) have a genetic component [i.e., are heritable; Andreola et al., 2020], both cross-sectional and the few available longitudinal studies suggest that these skills are not “set like plaster” after childhood but continue to change across the lifespan (Desjardins & Warnke, 2012; OECD, 2016; Paccagnella, 2016).

Second, although skill change during adulthood may involve both gains and losses, gains and losses typically occur at different ages. As studies using data from international large-scale comparisons such as the Programme for the International Assessment of Adult Competencies (PIAAC) show, the cross-sectional age profile of literacy and numeracy follows an inverted U-shape: Literacy and numeracy skills "increase" (cross-sectionally) with age throughout the second decade of life, peak at around an age of 30 years, and gradually "decline" thereafter (e.g., Gabrielsen & Lundetræ, 2014; OECD, 2016; Paccagnella, 2016; Podolskiy & Popov, 2014). The cross-sectional age differences in skills are substantial: On average across participating countries, older adults (aged 55–65 years) score about 30 scale points lower on the PIAAC literacy scale (the equivalent of 0.8 SD) than young adults aged 25–34 years (Paccagnella, 2016).

Interestingly, the age trajectories of literacy and numeracy in the above studies neither follow Cattell's ideal-typical paths of Gc nor Gf but are best described as a mixture of both. The reason might be that although literacy and numeracy constitute acquired skills that are sensitive to education and experience (like Cattell's Gc and Baltes's cognitive pragmatics), but they also depend on _Gf_-type abilities or cognitive mechanics such as processing visual stimuli and manipulating them in working memory. For comparison, longitudinal research on cognitive aging focusing on basic cognitive abilities found approximately linear declines in _Gf_-type cognitive abilities (e.g., processing speed, reasoning, memory) from early adulthood onward that accelerated in old adulthood (Salthouse, 2019)_—_with the exception of increases with age in vocabulary. Vocabulary, of course, is an acquired skill, which is closer to Cattell's conception of Gc than to Gf and an essential part of literacy skills.

Third, beyond average age trends, there are additional individual differences and group differences in skill change. For example, using cross-sectional PIAAC 2012 data, Paccagnella (2016) compared age differences in literacy and numeracy skills among adults with different levels of educational qualification (i.e., primary, secondary, and tertiary or above). He found that among those with the highest educational qualification, the apparent decreases of skills with increasing age were more pronounced than among those with lower qualifications. By contrast, a recent longitudinal study found that higher education had a protective effect against declines in literacy over time after controlling for initial level of literacy skills (Wicht et al., 2020). In addition, while some studies suggest a tendency of female adults having relative strength in literacy and of male adults having relative strength in numeracy across the life course (e.g, Houtkoop & Jones, 1998; Satherley & Lawes, 2008), other studies show only a small to non-existent gender differences in adults' literacy skills (Solheim & Lundetræ, 2018). It remains unclear whether cross-sectional gender differences translate to differential change over time, and some studies did not find any pronounced gender differences in cross-sectional PIAAC skill profiles (OECD, 2016) or in skill change over time (Reder, 2009; Wicht et al., 2020). Another longitudinal study found women to start off at higher literacy levels but experience smaller proficiency gains over time compared to men (Wolf & Jenkins, 2014).

Although these prior studies have greatly advanced our knowledge about skill development in adulthood, the majority of these studies is based on large-scale cross-sectional surveys such as PIAAC and its predecessors (i.e., the International Adult Literacy Survey in 1994 and 1998 or the Adult Literacy and Life Skills Survey in 2003, 2006, and 2008); or on small-scale longitudinal studies based on selective samples such as the longitudinal study of adult learners (LSAL) that focuses on high-school dropouts in the US (Reder, 2009) or adults in basic skills programs in the UK (e.g., Wolf & Jenkins, 2014). Cross-sectional studies can be advantageous in that they do not suffer from selective attrition of low-skilled or lower-educated individuals, as most longitudinal assessment surveys do (Martin et al., 2020). Moreover, cross-sectional studies are unbiased from retest artifacts which plague some longitudinal surveys, especially ones that apply traditional designs in which the exact same items are administered repeatedly (Salthouse, 2019). At the same time, cross-sectional studies of age differences are limited in that they are unable to disentangle age-related changes from cohort effects. That is, they are unable to ascertain whether the putative age differences are due to age-related changes or stem from preexisting differences in skills between different birth cohorts (which may alreacy have arisen in childhood or adolescence). Small-scale longitudinal studies based on selective samples, meanwhile, are limited in that their findings may not generalize to the population as a whole. Moreover, by their very nature, these studies cover some subgroups (e.g., high-school dropouts) in a certain life stage but not other subgroups and life stages that may be of interest to policymakers and practitioners. Also, compared to literacy, the life-span development of numeracy has received much less attention in prior research, despite the frequently encountered argument in the literature that numeracy skills are gaining in importance on today's labor markets (e.g., Gal et al., 2020; Gauly et al., 2020). In order to overcome the limitations of cross-sectional and small-scale longitudinal designs, repeated measures of literacy and numeracy skills are needed. Such data have long been in short supply. Until very recently, there were no data sources available internationally that combined the following desirable features that would allow for complete and robust answers to questions surrounding age-related changes in skills during adulthood: A large and non-selective sample; objective, and high-quality skill assessments; and a repeated measures design.

Here we leverage the unique3 potential of two recent German large-scale assessment surveys that do meet the above criteria: PIAAC-longitudinal (PIAAC-L), a follow-up to the 2012 Programme for the International Assessment of Adult Competencies (PIAAC) study in Germany; and Starting Cohort 6 from the National Educational Panel Study (NEPS). Both surveys offer repeated_—_and largely comparable_—_measures of adults' literacy and numeracy spaced three (PIAAC-L) to six years (NEPS) apart. Combining these data offers us the opportunity to analyze changes in literacy and numeracy skills overtime during adulthood, allowing us to present what appear to be the most comprehensive descriptive analyses of age-related changes of literacy and numeracy skills during adulthood to date based on a repeated-measures design.

With these data, we seek to answer two guiding questions about the stability and change of adults' literacy and numeracy skills. First, to what extent do these skills change across the three- to six-year periods covered by PIAAC-L and NEPS? Second, does the extent of age-related change differ across major sociodemographic groups?

Section snippets

PIAAC/PIAAC-L

Our first data source is the German PIAAC/PIAAC-L study [version 3–0-0; GESIS – Leibniz Institute for the Social Sciences et al., 2017]. PIAAC was conducted in 2012 and provides internationally comparable data on the skills of the working-age population (16–65 years) residing in private households in 24 countries (OECD, 2013b). Germany implemented a registry-based sampling design, in which respondents were randomly sampled from local population registers in randomly selected municipalities. A

Descriptive statistics on skills and skill change

Before reporting our two outcomes of focal interest–mean-level change and rank-order consistency—we first inspected descriptive statistics on skills and skill change in the original raw metric of the assessments (0–500 points in PIAAC; logits in NEPS). These are shown in Table 1 (based on unweighted sample data) and supplementary Table A1 (based on weighted data). These descriptive statistics show that the means (M) and standard deviations (SD) of both literacy and numeracy hardly changed

Discussion

Are literacy and numeracy skills impervious to change in adulthood—or do they continue to change? Does change involve gains or losses, and how are gains and losses distributed in the different segments of the adult population? To answer these questions, we analyzed data from two large-scale assessment surveys offering repeated measures of adults' literacy and numeracy skills spaced three (PIAAC-L) to six (NEPS) years apart. We considered two indicators of skill change: absolute change (based on

CRediT authorship contribution statement

The authors made the following contributions:

Funding information, data availability, and replication files

This work was supported by two grants to Clemens M. Lechner from the German Federal Ministry of Education and Research (BMBF), project “Risk and protective factors for the development of low literacy and numeracy in German adults” (Grant No. W143700A) and the German Research Foundation (DFG), project "Stability and Change in Adult Competencies: Patterns and Predictors of Literacy and Numeracy Development" (Grant No. LE 4001/1-1).

Data from the two studies used in this investigation are public

Cited by (14)

2023, International Journal of Educational Research
As immigrant groups and their children comprise a growing number of each country's population, questions of how to support their successful long-term integration into society are more important than ever (McHugh & Doxsee, 2018). Literacy is considered a foundational skill for successful school performance as well as civic, social, and economic participation (Borgonovi, Choi & Paccagnella, 2021; Lechner, Gauly, Miyamoto & Wicht, 2021; Windisch, 2016; Wrigley, Chen, White & Soroui, 2009). Alarmingly, multilingual immigrant learners1 have been substantially overrepresented in the low literacy segment of most societies (Grotlüschen, Mallows, Reder & Sabatini, 2016; Netten, Luyten, Droop & Verhoeven, 2014; OECD, 2015).

2022, Studies in Educational Evaluation
Bloom’s taxonomy was originally developed as a way to help instructors to achieve a range of educational objectives within their teaching practice. Based on analyses of PIAAC data, we know that a significant proportion of the adult population lacks proficiency in basic skills (Durda, Gauly, Buddeberg, Lechner, & Artelt, 2020; Lechner, Gauly, Miyamoto, & Wicht, 2021; Paccagnella, 2016). It is important that education scholars receive more in-depth information on the nature of the exact skills problems in order for to fine tune this area of research. View all citing articles on Scopus

© 2021 Elsevier Ltd. All rights reserved.