Mitigating wage penalties from education–occupation mismatches through workplace learning among manufacturing workers Open Access

Due to the expansion of higher education and rapidly changing industrial needs, the education–occupation mismatch has emerged as a critical issue linked to negative labor market outcomes. In general, the education–occupation mismatch can be understood in two ways: horizontal mismatch, which occurs when a worker, educated in one field (i.e. English), works in a different field (i.e. statistician) and vertical mismatch, which refers to the discrepancy between a worker’s education level and the job’s educational requirements (Sloane, 2003). Notably, horizontal mismatch – also known as field-of-study mismatch – is a common phenomenon but has received less attention than vertical mismatch (Somers et al., 2019). An Organization of Economic Co-operation and Development (OECD) study using the Program of International Assessment of Adult Competencies (PIAAC) found that approximately 40% of workers are horizontally mismatched across countries: approximately 50% in Korea, 45.3% in Japan, 45% in the USA and 26.4% in Germany (Montt, 2015).

Both horizontal and vertical mismatches are detrimental not only to individuals but also to the economy. Empirical evidence indicates that the education–occupation mismatch is associated with lower wages (Choi et al., 2020; Kim et al., 2016; Montt, 2017; Robst, 2007; Tran et al., 2025), lower job satisfaction (Hur et al., 2019; Wen et al., 2023), increased turnover intentions (Chavadi et al., 2022; Choi and Hur, 2020; Wolbers, 2003) and higher job quitting (Wen et al., 2023). Despite these challenges, little attention has been paid to how wage penalties associated with the education–occupation mismatch can be mitigated. In an ideal scenario, firms maximize productivity by recruiting workers whose skills perfectly match job requirements. In practice, however, mismatches frequently arise due to information asymmetries among job seekers, wage ceilings in certain fields and segmented labor market structures (Kim et al., 2016). These challenges are particularly acute for early-career workers with fewer employment opportunities (Sevilla and Farías, 2020; Tran et al., 2025).

To date, most existing studies have concentrated on providing empirical evidence of the detrimental effects of education–job mismatch, while paying little attention to practical strategies for mitigating these effects. Among the few exceptions, Robst (2007) found that the wage penalty associated with field-of-study mismatch diminishes over time as workers gain tenure, suggesting that human capital development can function as a mitigating mechanism. From a psychological perspective, Chavadi et al. (2022) showed that job satisfaction among mismatched workers can reduce their turnover intentions. However, these studies still offer limited insights into specific practices and interventions through which the negative effect of mismatch on productivity can be alleviated.

This study addresses that gap by examining how workplace learning can help workers experiencing horizontal mismatch reduce their wage penalties from an organizational practice perspective. Workplace learning is known to enhance workers’ job satisfaction, engagement and productivity (Decius et al., 2021; Rowden, 2002). Building on this evidence, this study posits that workplace learning enables mismatched workers to acquire job-specific knowledge and skills, thereby improving productivity and mitigating the mismatch-relevant wage penalties. The present study specifically focuses on the manufacturing sector, which has suffered from labor shortages (Freifeld, 2022) and a higher likelihood of education and skill mismatches (Weaver and Osterman, 2017). Findings using cross-national PIAAC data illustrate empirical evidence for the mitigating effect of workplace learning on the relationship between horizontal mismatch and earnings. In addition, subgroup analyses by workers’ role and firm size enhance our understanding of which workplace learning strategies are most effective for mismatched workers across various settings.

Multiple theoretical perspectives, such as human capital theory, career mobility theory and assignment theory, offer complementary insights into the effects of education–occupation mismatch (Hartog, 2000). First, human capital theory emphasizes the positive wage returns associated with investments in education and training (Becker, 1962) and predicts that wages may reflect productivity even in the short run (Kim et al., 2016). From this perspective, the wage penalties observed among overeducated workers can partly result from biased estimation due to imperfect measurement of human capital (McGuinness, 2006). Still, this theory does not necessarily rule out the potential that overeducation has a genuine negative effect on wages (Kim et al., 2016). Education–job mismatch is a temporary phenomenon (Hartog, 2000), as firms and workers gradually adjust through experience accumulation and job-embedded training (Kim et al., 2016). Although temporary, mismatch can impose substantial costs, especially for temporary workers who often receive fewer firm-sponsored training opportunities (Somers et al., 2019) and for early-career workers, who may initially appear underpaid relative to their education until their full productivity potential is realized (Kim et al., 2016).

Career mobility theory further formalizes this scenario (Hersch, 1991). Under this view, the education–occupation mismatch can emerge during the early stages of a worker’s career, particularly among young graduates and entry-level workers (Tran et al., 2025). Young workers may initially accept positions for which they are overqualified or in unrelated fields because such roles are more accessible. As they accumulate experience and skills, these workers can transition to higher-paying, better-matched positions through intra- and inter-firm mobility. From this perspective, horizontal mismatch may represent a short-term compromise, accepting lower wages initially for the potential of higher earnings later (Kim et al., 2016).

Conversely, assignment theory emphasizes occupational characteristics, such as job requirements, productivity ceilings or segmented labor market structures, in predicting worker productivity (Sattinger, 1993). From this perspective, individuals are assigned to specific jobs within an unequal labor market. Due to imperfect information, search frictions and high job search costs, workers may remain in positions that do not fully match their qualifications, leading to persistent wage penalties (Kim et al., 2016).

Overall, while all three theories agree that horizontal mismatches lead to wage penalties and reduced productivity, they differ in their views on the duration of these effects. This theoretical foundation has motivated empirical studies examining wage penalties associated with horizontal mismatch across diverse contexts.

Research consistently shows that horizontally mismatched workers experience wage penalties (Choi et al., 2020; Kim et al., 2016; Montt, 2017; Robst, 2007; Wolbers, 2003). Nordin et al. (2010) using Swedish data, reported that income penalties for horizontal mismatch exceed 30%, though this may be overestimated because vertical mismatch was not controlled. Accounting for overqualification, Kim et al. (2016) found strong wage penalties among Korean college graduates, particularly in the lower and middle earnings distribution.

Horizontal mismatch is more prevalent among private-sector workers, part-time or temporary employees and those in small and medium-sized enterprises (SMEs) (Hwang, 2018; Wolbers, 2003). Even with general skill transferability, horizontal mismatch persists (Robst, 2007). Montt (2017) showed that, even after considering skill transferability and field saturation, mismatched workers continue to face wage penalties, which grow when combined with overqualification. Albert et al. (2023) further observed that horizontal mismatches persist longer than vertical mismatches in certain situations, partly because mismatched workers are less likely to secure permanent contracts, making corrective career movements more difficult:

Workplace learning enables workers to acquire job-related knowledge, enhance performance and support organizational success. It encompasses both formal training opportunities and informal learning that occurs naturally in the work environment (Coetzer et al., 2020). Formal learning is organized, intentional and institutionally sponsored within a set timeframe, such as workshops and seminars, onboarding programs and professional certifications, whereas informal learning is unstructured and unintentional, occurring through social interaction, reading and observation (Eraut, 2000). Depending on research objectives, workplace learning is conceptualized as encompassing formal, informal and incidental learning (Shah et al., 2023; Watkins and Marsick, 1992). When defined more specifically, it is also described as formal training programs and informal social interactions that develop skills, attitudes and knowledge (Rowden, 2007). Together, these perspectives reflect its multifaceted nature, integrating both structured and unstructured learning (McCormack, 2000).

Although workers’ learning and training experiences vary across occupations (Harteis et al., 2015) and industry sectors (Jeong, 2026), and training systems differ by country-level labor market structure and institutions (Ehlert, 2020), the literature broadly agrees that workers benefit from a diverse range of workplace learning practices. These include training workshops, coaching, on-the-job training (OJT), blended learning, collaborative learning, microlearning and online learning (Coetzer, 2006; Decius et al., 2021; Rowden, 2002; Wallo et al., 2022). In the postpandemic era, technological advances and the expansion of remote and hybrid work arrangements have further reshaped the inherently social nature of learning at work by increasing reliance on virtual environments, flexible work options, and digital collaboration tools (Lane et al., 2024; Zajac et al., 2022). While these changes have the potential to reduce top-down communication, promote more inclusive interactions, support knowledge sharing and foster a culture of lifelong learning (Vallo Hult and Byström, 2022; Zajac et al., 2022), challenges still remain. In particular, many organizations have reduced investments in training and narrowing learning opportunities primarily to online instruction, increasingly shifting responsibility to workers to find their own learning solutions and independently develop their skills (Beier et al., 2025; Hughes et al., 2020).

While there is a growing body of literature that supports the benefits of workplace learning, empirical studies rarely examine whether workplace learning moderates mismatch-related wage penalties. Among the limited evidence, Nordin et al. (2010) found that the wage returns to work experience (tenure) were higher for mismatched men than for well-matched men, while no significant difference was found between mismatched and well-matched women. This finding suggests that human capital accumulated through work experience may substitute education-specific skills acquired in formal settings. Hur et al. (2019) reported significant wage penalties for horizontal mismatch, especially for minority women and further suggested – but did not empirically test – that workplace training, such as mentoring, formal education or OJT, could mitigate negative workforce outcomes.

Overall, workplace learning is expected to help mismatched workers to acquire field-specific skills valued by employers, enhance their productivity and improve career prospects, thereby mitigating wage penalties:

This study further proposes that the extent to which workplace learning mitigates wage penalties associated with horizontal mismatches depends on workplace characteristics. Differences in management systems and learning environments across manufacturing firms (Coetzer, 2006) create unique contexts for the role of workplace learning for mismatched workers.

The mitigating effect of workplace learning may vary by workers’ roles. Horizontal mismatch is more common among subordinate workers, since it often occurs early in careers or entry-level positions (Albert et al., 2023; Nordin et al., 2010). However, the negative wage impact may be stronger for managers, given steeper salary progression and the importance of field-specific knowledge, mentoring and supervisory skills in leadership roles (Northouse, 2021). Managerial positions demand close alignment between educational background and job responsibilities to demonstrate both productivity and leadership potential. For mismatched managers, a misaligned field-of-study background may weaken the signal of expertise and limit wage growth or promotion opportunities.

In this context, the mitigating role of workplace learning may vary by workers’ positions. Because managerial and nonmanagerial positions demand different competencies (executive, interpersonal, informational and decisional responsibilities versus more technical or task-specific skills) (Northouse, 2021; Yamazaki et al., 2018), their learning opportunities also differ. Managers often engage in leadership development and cross-functional training, while nonmanagerial workers rely more on OJT, skill-specific training and team-based learning (Wallo et al., 2022).

Given these distinctions, workplace learning may differentially mitigate mismatch-related penalties for managers and subordinates. Subgroup analyses by managerial status are conducted to test this expectation:

The study also postulates that the mitigating effect of workplace learning varies by firm size. For example, SMEs typically pay lower wages, face wage floors, operate in smaller markets and experience stronger cost pressures that limit pay increases, compared to large firms (Rowden, 2002; Wursten and Reich, 2023). They also have fewer resources for extensive recruitment, experience flatter structures with fewer advancement opportunities and may struggle to attract workers with strong education–job alignment (Coraggio et al., 2022). Empirical studies consistently show higher horizontal mismatch prevalence in SMEs than in large firms (Hwang, 2018; Montt, 2015).

Workplace learning opportunities also differ by firm size. SMEs often lack resources for formal training systems (Coetzer, 2006; Decius et al., 2021), leading employees to rely on flexible, informal learning, such as OJT, peer learning and work-integrated learning (Shah et al., 2023; Tamm, 2018), which typically requires the immediate application of acquired skills (Jeong et al., 2018). Large firms, in contrast, maintain structured training systems, including formal onboarding, in-house programs and specialized training departments that use advanced tools, such as e-learning and virtual reality (Lee et al., 2015).

These contextual differences in wage structures, mismatch prevalence and learning opportunities suggest that workplace learning’s mitigating effect on mismatch-related penalties will vary by firm size. Subgroup analyses by firm size access this expectation:

This study uses data from the OECD’s Program for the International Assessment of Adult Competencies (PIAAC), which provides comprehensive information on sociodemographic characteristics, education and training, skill use and job characteristics (OECD, 2010). The survey was administered to approximately 5,000 adults aged 16–65 years across 39 countries from 2012 to 2017.

The analytic sample includes wage-employed workers in the manufacturing sector (International Standard Industrial Classification (ISIC) “C. Manufacturing”) who completed a field-specific program at the upper-secondary level or higher (International Standard Classification of Education (ISCED) 3 or above). Self-employed workers are excluded due to their distinct work characteristics and wage structures. The analysis uses PIAAC public-use data which includes hourly wages, field-of-study programs and three-digit occupational codes (International Standard Classification of Occupations (ISCO)-08). Because variable availability differs slightly across countries, listwise deletion was used to ensure comparability and consistent model estimation.

The final data set comprises 6,939 individuals from 23 countries: Belgium, Chile, Cyprus, the Czech Republic, Denmark, Ecuador, France, Greece, Israel, Italy, Japan, Kazakhstan, Korea, Lithuania, Mexico, the Netherlands, New Zealand, Norway, the Russian Federation, the Slovak Republic, Slovenia, Spain, Turkey and the UK. The sample consists of 68.1% men and 31.8% women: adults aged 16–24 (11.3%), 25–34 (27.3%), 35–44 (27.3%), 45–54 (22.9%) and 55 plus (11.2%). Regarding education, 56.7% hold a high school diploma or below, 5.6% a postsecondary degree, 27.5% a bachelor’s degree and 10.2% a master’s degree or higher.  Appendix 1 provides country-level sample sizes and mismatch rates.

Country sample sizes range from 65 to 613. Although countries such as Greece (n = 65) and Ecuador (n = 93) have smaller samples, the analysis draws on the pooled cross-national data set, where country effects are controlled but not the focus of interpretation. Variation in country sample sizes does not inherently bias pooled OLS estimates under standard assumption (Wooldridge, 2013). Provided that model diagnostics were checked and properly addressed (see the Analytical Strategies section), these small subsamples are unlikely to substantially affect the validity of the overall pooled results (Gujarati, 2009). In addition, sensitivity analyses excluding these smallest countries remain consistent with the main findings (Mayen et al., 2024).

This study focuses exclusively on the manufacturing sector. While prior studies typically pool workers across industries to examine mismatch-related wage penalties (Choi et al., 2020; Montt, 2017; Nordin et al., 2010), this study investigates how workplace learning moderates these penalties within a single sector, where conditions may differ substantially.

Manufacturing firms, particularly SMEs, often struggle to fill positions despite high job openings. For example, in South Korea, manufacturing remains central to the economy, contributing over 25% of GDP and 90% of exports (International Trade Administration, 2023). SMEs comprise 97.4% of manufacturing firms (Korea Ministry of SMEs and Startups, 2024) and face persistent workforce shortages: a 2025 survey of 589 SMEs reported that 28.9% experienced labor gaps, and nearly 25% anticipated even greater hiring difficulties (Noh, 2025). Such shortages heighten the likelihood of education and skill mismatches (Weaver and Osterman, 2017) and highlight the importance of workplace learning as a response to skilled labor scarcity (Freifeld, 2022).

Workplace learning in manufacturing often emphasizes hands-on, practical and role-specific protocols related to machine operation, technology use and quality control (Mori, 2023). A substantial portion is also compliance-oriented, covering safety, environmental regulations and emergency preparedness, which is essential, but it may not always build new skills. Although PIAAC does not distinguish training types, it nonetheless enables an examination of how workplace learning may mitigate wage penalties for horizontally mismatched workers and offers practical insights for the manufacturing sector.

Key measures include wages, horizontal mismatches and workplace learning.

The outcome variable is logged hourly wages, including bonuses and adjusted for purchasing power parity in US dollars to ensure cross-country comparability. To reduce outlier influence, values below the 1st and above the 99th percentile were winsorized prior to logarithmic transformation.

Horizontal mismatch compares respondents’ highest field-of-study with their three-digit occupational code (ISCO-08), following Montt’s (2015) crosswalk. For example, the “teacher training and education science (1)” category corresponds to various occupations such as university, higher education, vocational, secondary, primary, early childhood and other teaching professionals (ISCO 231–235); sports and fitness workers (ISCO 342); and childcare workers and teachers’ aids (ISCO 531). Respondents are coded as well-matched (1) if their program aligns with the corresponding occupational codes; otherwise, they are coded as a mismatch (0).

Matching may occur in one-to-one or one-to-many cases. In addition to Montt (2015)'s crosswalk, this study also classifies field-specific managerial occupations (ISCO 121–143) as matched when the managerial role corresponds to the worker’s field-of-study. For instance, a field-of-study in engineering, manufacturing or construction matched with managerial occupations of manufacturing, construction and distribution (ISCO 132). Reflecting career progression pathways, directors and chief executives (ISCO 112) are coded as matched regardless of their field of study, not necessarily from only a business major.

Workplace learning is measured using three PIAAC items (OECD, 2011). The main index, workplace learning overall (WLO), sums three binary indicators: open or distance learning (ODL), OJT and participation in Seminars/Workshops (B_Q12a, B_Q12c and B_Q12e), yielding a score ranging from zero to three. These items provide a snapshot of human resource investments by type of training and are considered one of the key determinants of skills acquisition and labor market outcomes (OECD, 2011). Accordingly, the composite index captures the breadth of workers’ engagement in learning activities, offering a parsimonious summary of overall learning exposure. At the same time, each workplace learning variable is modeled separately with its interaction term to facilitate detailed interpretation.

Specifically, ODL is measured by asking, “During the last 12 months, have you participated in courses conducted through open or distance education (B_Q12a)?” This captures courses delivered via postal correspondence or electronic media, connecting instructors with remote learners. OJT is measured by asking, “During the last 12 months, have you attended any organized sessions for OJT or training by supervisors or co-workers (B_Q12c)?” This includes planned training sessions, practical instruction or on-site experiences using standard work tools, which may include either general or job-specific guidance. Seminars/workshops is measured by asking, “During the last 12 months, have you participated in seminars or workshops (B_Q12e)?”

Control variables include demographic characteristics (gender, age group, highest education, field-of-study of highest education, parents’ highest education, immigrant status, numeracy scores and over-education status), work characteristics (firm size, tenure, tenure-squared, permanent contract status and managerial positions) and 22 country dummies. Firms with fewer than 250 employees were coded as SMEs and those with 250 or more as larger firms (OECD, 2024). Managerial positions refer to workers who supervise or manage others.

Following prior studies (Choi and Hur, 2020; Kim et al., 2016; Montt, 2017), Mincer’s earnings equation was used in the analysis:

where $Wi$ denotes the log of hourly wages and $HMi$ is an indicator for horizontal mismatch. The moderator, workplace learning (⁠$WLi$⁠) is entered as either the composite WLO (Model 1–1) or its components – ODL, OJT and Seminar/Workshops (Models 1–2 to 1–4). The interaction term $WLi*HMi$ captures whether workplace learning moderates mismatch-related wage penalties. Because the dependent variable is log-transformed, coefficients can be interpreted as approximate percentage changes in wages $(%Δwage=(eb−1)*100)$ for a one-unit change in the predictor.

Covariate $Xi$ includes demographic characteristics. In addition, overeducation (or overqualification) and individual proficiency (numeracy) are controlled to avoid overestimating horizontal mismatch effects. $Yi$ covers job-related characteristics, while $Zi$ represents the country dummy variable. Although these controls reduce bias, unobserved factors may remain and the cross-sectional PIAAC design limits causal inference. Subgroup analyses by firm size and workers’ role test contextual differences in the moderating role of workplace learning.

Diagnostic checks indicate that multicollinearity is not a concern (mean VIF = 3.82). Higher VIFs for age-squared and some categorical variables are expected by construction and did not affect results when these terms were removed or categories collapsed. The Breusch–Pagan test indicated heteroskedasticity [χ²(1) = 123.23, p < 0.001], addressed by log-transforming wages and using jackknife robust standard errors via the repest package in Stata, which incorporates replicate weights and plausible values for numeracy (Avvisati and Keslair, 2024).

In the analysis sample, approximately 41.9% of manufacturing workers are horizontally mismatched. Compared with workers in other industries within PIAAC data, this rate is higher than in construction (29.80%) and utilities (35.1%), but lower than transportation and storage (44.7%) and accommodation/food service (57.7%).

Table 1 presents mismatch patterns across key variables. Horizontal mismatch is more common in SMEs (44.0%) than in larger firms (39.8%). Nonmanagerial workers (44.2%) exhibit higher mismatch rates than managers (39.1%). Workers who engaged in workplace learning – ODL, OJT, seminars/workshops – generally exhibit lower mismatch rates than those without such experiences. The largest gap appears in OJT: 38.8% of OJT participants are mismatched versus 45% of nonparticipants. Well-matched workers also report higher WLO scores than mismatched workers (0.739 vs 0.656; $t=4.046, p<.001,Cohen′s d=0.10$⁠). Similarly, well-matched workers earn more than mismatched workers (⁠$t=10.863, p<.001, d=0.26$⁠). All effect sizes fall within Cohen’s (1992) “small” range (d = 0.10 to 0.26).  Appendix 2 presents correlations among the variables.

Table 2 summarizes mismatch patterns by firm size and managerial status. Wage disparities between well-matched and mismatched workers are significant in both SMEs (⁠$t=9.15, p<.001, d=0.27$⁠) and large enterprises (⁠$t=5.26, p<.001, d=0.22$⁠). In SMEs, well-matched workers show higher participation in WLO (⁠$t=3.33, p<.01,d=0.10)$ and in OJT (⁠$t=4.68, p<.001, d=0.14$⁠) than mismatched workers, but these differences do not appear in large enterprises. It is important to note that these figures reflect descriptive associations from cross-sectional data and should not be interpreted causally as evidence that mismatch restricts training access.

Regarding managerial status, wage gaps between well-matched and mismatched workers are significant among both subordinates (⁠$t=10.13, p<.001,d=0.29$⁠) and managers (⁠$t=3.24, p<.001,d=0.14$⁠). Among subordinates, well-matched workers report higher WLO and OJT participation than mismatched workers (⁠$t=4.99, p<.001, d=0.15$⁠; $t=5.62, p<.001, d=0.16$⁠, respectively), while differences are minimal among managers. Effect sizes range from −0.06 to 0.29 (small).

H1 tests whether horizontally mismatched workers experience greater wage penalties than well-matched workers (see Table 3;  Appendix 3 for full results). Overall, the results indicate a consistent negative association between horizontal mismatch and wages.

In the WLO model (Model 1–1), horizontal mismatch has a negative effect on wages (⁠$b=−0.074, p<0.05$⁠), corresponding to approximately a 7.1% reduction in wages $(7.1%=(e−0.074−1)*100)$⁠. Similar effects appear in the OJT model (Model 1–3; $b=−0.061,p<0.05$⁠) and seminars/workshops model (Model 1–4; $b=−0.059, p<0.05$⁠), even after controlling for overqualification. The only exception was found in the ODL model (Model 1–2; $b=−0.031, p=0.191$⁠). Taken together, these findings are consistent with H1.

H2 examines whether workplace learning moderates the relationship between horizontal mismatch and wages. First, all learning variables (WLO, ODL, OJT and seminars/workshops) show positive main effects on wages (Models 1–1 through 1–4).

Regarding the interaction effect, workplace learning shows statistically significant interactions with horizontal mismatch. In Model 1–1, WLO positively moderates the association between mismatch and wages (⁠$b=0.067, p<0.01$⁠). When simple effects were calculated, without WLO, mismatched workers earn 7.1% less than well-matched workers$(b=−0.074, p<0.01; 7.1%=(e−0.074−1)*100)$⁠. With WLO ( = 1), however, holding covariates constant, mismatched workers ( = 1) have a wage premium$(b=0.085=−0.074 +0.092 +0.067)$⁠, corresponding to an 8.9% increase $(8.9%=(e0.085−1)*100)$ relative to the reference group (well-matched workers without WLO).

Similarly, OJT (Model 1–3; $b=0.091, p<0.05$⁠) and seminars/workshops (Model 1–4; $b=0.149, p<0.01$⁠) show positive moderating effects. Simple effects show that mismatched workers without seminars/workshops earn about 5.7% less relative to the baseline group (well-matched workers without seminars or workshops [SW]), whereas those with seminars/workshops have nearly a 20.2% wage premium. For OJT, mismatched workers earn approximately 5.9% less without OJT, but those with OJT receive an 11.2% wage premium relative to the reference group. However, ODL does not have the moderating effect (Model 1–2; $b=0.050, p=0.413$⁠). Overall, these results suggest that workplace learning – WLO, OJT and seminars/workshops – mitigates the wage penalties associated with horizontal mismatches among manufacturing workers, consistent with H2.

Subgroup analyses test H3 using samples of managers and subordinates (see Table 4;  Appendix 4). Mismatched managers experience a significant wage penalty (Model 2–1; $b=−0.195,p<0.01$⁠), whereas mismatched subordinates did not (Model 3–1; $b=−0.023, p=0.610$⁠).

Among managers, WLO does not directly affect wages (Model 2–1; $b=0.045, p>0.05$⁠), but significantly moderates mismatch penalties (⁠$b=0.105, p<0.05$⁠). Simple effects show that mismatched managers without WLO earn 17.7% less than the reference group (well-matched managers without WLO), whereas those with WLO earn only 4.4% less, indicating a smaller wage gap associated with horizontal mismatch among managers with WLO. Similarly, OJT (Model 2–3; $b=0.187, p<0.05$⁠) and seminars/workshops (Model 2–4; $b=0.226, p<0.01$⁠) significantly mitigate wage penalties, while ODL does not (Model 2–2; $b=0.060, p=0.143$⁠).

Among subordinates, WLO positively affects wages (Model 3–1; $b=0.110, p<0.001$⁠). However, as the mismatch itself is not significantly associated with wages in this group (⁠$b=−0.023, p=0.610$⁠), the interaction term is also not significant (⁠$b=0.056, p=0.100$⁠), which is consistent across Models 3–2 to 3–4.

Overall, these findings are consistent with H3, indicating that the mitigating effect of workplace learning on wage penalties associated with horizontal mismatch varies by managerial status, with larger estimated interactions among managers.

Subgroup analyses by firm size test H4 (see Table 5;  Appendix 5). In SMEs, mismatched workers experience wage penalties (Model 4–1; $b=−0.104, p<0.05$⁠), significant in the OJT (Model 4–3; $b=−0.117, p<0.01$⁠) and seminars/workshops (Model 4–4; $b=−0.070, p<0.1$⁠) models. However, horizontally mismatched workers in large enterprises do not face such penalties (Models 5–1 to 5–4).

Among SME workers, WLO and its interaction are significant (Model 4–1; $b=0.083, p<0.001$⁠; $b=0.090, p<0.05$⁠). Mismatched SME workers without WLO earn 9.9% less than the reference group (well-matched SME workers without WLO), whereas those with WLO earn 7.1% more. OJT also shows a significant moderating effect (Model 4–3; $b=0.203, p<0.001$⁠): mismatched SME workers without OJT earn 11% less, while those with OJT are estimated to have a 14.1% wage premium. However, the interaction effects for ODL (Model 4–2; $b=0.019, p=0.808$⁠) and seminars/workshops (Model 4–4; $b=0.107, p=0.196$⁠) are not significant.

In large firms (Model 5), WLO, OJT and seminars/workshops positively affect wages. However, the WLO interaction with mismatch is not significant (Model 5–1; $b=0.031, p=0.339)$⁠, as horizontal mismatch itself has no significant effect (⁠$b=−0.026, p=0.627$⁠). Only the seminars/workshops interaction is significant (Model 5–4; $b=0.193, p<0.01$⁠).

These findings are consistent with H4, indicating that the mitigating effect of workplace learning on wage penalties associated with horizontal mismatch varies between SMEs and large enterprises, with strong estimated interactions among workers in SMEs.

Using PIAAC data from 23 countries, this study examined whether workplace learning mitigates wage penalties associated with horizontal mismatches among manufacturing workers and how these effects vary by worker role and firm size. The discussion and conclusion are presented below.

First, horizontally mismatched manufacturing workers experience wage penalties, even after controlling for overqualification, demographic and job characteristics and country-fixed effects. This is consistent with prior evidence for college graduates in the U.S. (Hur et al., 2019) and Korea (Kim et al., 2016), workers in Thailand (Serikbayeva and Abdulla, 2022) and workers across industries using PIAAC data (Choi et al., 2020; Montt, 2017). The results align with human capital, career mobility and assignment theories, which all predict lower earnings for workers whose jobs are mismatched with their field-of-study (Montt, 2017; Robst, 2007; Wolbers, 2003).

Second, WLO, OJT and seminars/workshops significantly reduce wage penalties for mismatched workers, complementing prior studies that highlight training and learning as a key mechanism for alleviating mismatch-related wage penalties (Hur et al., 2019; Nordin et al., 2010). In line with human capital theory, continuous training and learning foster occupation-specific skill development and enhance productivity (Kim et al., 2016). Education and training implicitly and explicitly yield stronger skill development (Ferreira et al., 2016) and enable mismatched workers to strengthen their competencies, thereby enhancing productivity and wages (Grip and Andries, 2015).

By contrast, Open and Distance Learning (ODL) does not significantly moderate mismatch-related wage penalties. Only about 10% of workers in the analytic sample engaged in ODL, compared with around 40% in OJT and 21% in Seminars/Workshops. During the 2012–2017 survey period, MOOCs and other online platforms expanded rapidly, but employer recognition of these credentials was often limited due to concerns about their quality, pedagogical rigor and limited face-to-face interaction (Linardopoulos, 2012; Tabatabaei et al., 2014). This underscores the need for future studies using more recent post-pandemic data to examine the evolving role and recognition of ODL in mitigating horizontal mismatch. Indeed, short-form online training, virtual reality training and digital collaboration tools enable dynamic environments for workers’ re-skilling and up-skilling (Bennett and McWhorter, 2021; Lane et al., 2024). Emerging technologies in smart manufacturing, including artificial interlligence, advanced analytics and human–robot collaborations (Rai et al., 2021; Wang et al., 2024) are also reshaping which competencies are valued and how they are developed (Vallo Hult and Byström, 2022), warranting further research on their implications for mismatch and workplace learning.

Third, the mitigating effect of workplace learning on mismatch-related wage penalties varies by managerial status. Horizontal mismatches are frequent among nonmanagerial workers, consistent with career mobility theory’s emphasis on early-career mismatches (Albert et al., 2023). Noteworthy is that mismatch-related wage penalties are pronounced among the managers' group. Managerial positions often involve steeper salary growth, greater field-specific expertise and more rigorous evaluations of leadership qualifications (Northouse, 2021). Persistent wage penalties among mismatched managers are consistent with assignment theory’s view that education–occupation mismatches can reflect long-term labor market inefficiencies (Kim et al., 2016; Sattinger, 1993).

Despite this persistence, workplace learning, particularly OJT and seminars/workshops, substantially reduces mismatch-related wage penalties for managers. Managerial roles often receive more comprehensive and targeted training, such as leadership development programs. These opportunities help mismatched managers build both field-specific and supervisory competencies, enabling them to narrow wage gaps relative to well-matched managers.

Finally, the mitigating effect of workplace learning on mismatch-relevant wage penalties varies by firm size. The results indicate that wage penalties associated with horizontal mismatch are evident among workers in SMEs but not among those in large firms. Several factors likely contribute: SMEs often face resource constraints that limit recruitment of workers with highly relevant experience (Coraggio et al., 2022); job seekers tend to prefer large firms offering better career prospects and wage growth (Araújo and Carneiro, 2023); and many SMEs operate in smaller or non-urban communities with lower costs of living (Rowden, 2002) and typically are more vulnerable to cost pressures and pay lower wages (Wursten and Reich, 2023). As shown in the results, within SMEs, OJT significantly mitigates wage penalties for horizontal mismatch, reflecting that SMEs’ flexible and informal learning environments help workers improve field-specific skills tailored to local production processes (Long et al., 2000; Tamm, 2018). In large firms, mismatched workers mostly benefit from seminars or workshops embedded in formal training systems, which also foster self-directed and reflective learning (Lee et al., 2015).

The findings suggest several implications for policy and practice: First, firms should provide tailored training and learning opportunities for mismatched workers, considering firm size and worker roles. Ferreira et al. (2016) demonstrate that formal training, informal learning and their complementarity substantially enhance skill development. This study indicates that mismatched workers benefit particularly from OJT and seminars/workshops. Organizations can strengthen workplace learning directly through performance- and development-oriented interventions and indirectly by cultivating environments that encourage knowledge sharing, skill application and a learning culture (Coetzer et al., 2019; Wallo et al., 2022). This underscores the dual importance of individual initiative and organizational support.

Managers and experienced co-workers in SMEs can act as learning facilitators, using effective communication and coaching to support entry-level and mismatched workers (Coetzer, 2006). SMEs can leverage flexible and informal learning approaches through OJT, microlearning and peer learning to provide intensive, context-specific development opportunities. Mismatched managers may benefit from more formal programs (e.g. leadership and professional development or certification courses) to gain employer recognition and enhance supervisory and field-specific expertise.

Second, the study highlights the need for adaptive partnerships between educational institutions and industry that can respond continuously to shifting labor market conditions. Rather than one-time alignment exercises, adaptive partnerships emphasize sustained collaboration and feedback (Lutchen, 2024). Advisory boards that include diverse employers can regularly inform curricular design, facility investments and work-based learning opportunities. Teaching and learning strategies can incorporate authentic, industry-based experiences such as “learning factory” models, industry-sponsored design projects and cocurricular workshops led by practitioners (Lamancusa et al., 1997; Murray et al., 2020). University-industry partnership centers can coordinate guest lectures, applied projects, internships and mentoring, while career seminars, employability workshops, capstone projects and graduate placement programs can be embedded within these frameworks to strengthen school-to-work transitions (Lutchen, 2024).

Third, workforce development policy should provide stronger institutional support for mismatched workers. Education–occupation mismatches are particularly common among immigrants, women or temporary and young workers (Albert et al., 2023; Choi and Hur, 2020). Many countries operate active labor market programs (ALMP) offering job search assistance, matching systems and rehabilitation programs, particularly for marginalized groups. Yet ALMP spending has declined, averaging only 0.6% of GDP among OECD countries in 2021 (OECD, 2021) and firms, particularly SMEs, remain reluctant to invest in training workers with low education levels (Bonvillian and Sarma, 2021). Publicly funded or co-funded strategies targeting mismatched workers are therefore crucial. Governments can expand the training vouchers, certification initiatives and individualized training plans to help workers acquire industry-recognized skills and credentials, while improving public employment services by enhancing alignment between job placements and workers’ skills and aspirations.

This study has several key limitations. First, its cross-sectional design based on observational data does not permit causal inference; unmeasured confounders may partly explain the observed relationship between mismatch, workplace learning and wage. Cross-national differences in education systems and labor market structures may influence the magnitude of wage penalties associated with horizontal mismatch, suggesting the value of multilevel modeling to examine contextual moderation.

Moreover, the binary measures of horizontal mismatch and workplace learning may obscure underlying heterogeneity. PIAAC does not capture nuanced subjective assessments of mismatch (i.e. “somewhat” vs “closely” matched) and has relatively low response rates (20%–40%) for items related to training time or expenditures, which made this study use binary indicators. Future studies should incorporate richer measures of both mismatch and workplace learning, including the contexts in which workplace learning occurs. Future studies can also examine the mitigating effects of workplace learning in other sectors (services, information technology, etc.) thereby identifying strategies that support mismatched workers in diverse labor market contexts.

Portion of this study was presented at 2024 Academy of Human Resource Development International Conference and 2024 Assoociation of Career and Technical Education Research Conference. The author sincerely appreciate the participants for their helpful comments and discussions. All errors are mine.

Appendices can be found in the online repository at Link to the cited article.