A Unified Validation Framework for Assessing Green Skills Implementation from Basic to Vocational Education: A Measurement Invariance Approach

2.1 Conceptualization of green skills

Green skills have evolved from a narrow focus on environmental knowledge toward a multidimensional construct encompassing cognitive, behavioral, and technical competencies required for sustainable development (Vona et al., 2015; Auktor, 2020). Recent literature conceptualizes green skills along a continuum from foundational environmental awareness to advanced technical competencies embedded in sector-specific practices (Wegenberger & Ponocny, 2025). This progression reflects the increasing complexity of sustainability challenges across economic sectors and the rising expectation that workers possess capabilities to adapt to green technologies, support eco-efficient production processes, and engage in environmentally responsible behaviours (Ristanović et al., 2025; Cao et al., 2025).

Albertz and Pilz (2025) further expand this conceptualization through a comprehensive literature review of green vocational education, identifying shared themes across institutional, geographic, and pedagogical contexts. Their analysis shows that green skills development is shaped by diverse actors—including policymakers, schools, industries, and communities—and that curricular alignment plays a vital role in ensuring relevance to labour market demands. These frameworks highlight the importance of embedding green skills across educational systems rather than limiting them to vocational education and training (VET).

2.2 Green skills across educational levels: From foundational to technical competencies

A growing body of research emphasizes the developmental nature of green skills, where competencies emerge gradually from early schooling and become increasingly specialized in vocational settings. In basic education, students develop fundamental environmental attitudes, ecological awareness, and sustainable habits that form the foundation for more advanced competencies (Ismail et al., 2024; Sukri et al., 2022). These early-stage competencies reflect what Wegenberger and Ponocny (2025) describe as foundational competences, emphasizing environmental literacy and personal responsibility.

In lower secondary education, students’ exposure to scientific inquiry, sustainability projects, and critical thinking activities supports the development of transitional competences. These competencies involve analyzing environmental issues, evaluating human–environment interactions, and participating in real-world sustainability initiatives. The shift from awareness to action is crucial in bridging basic education with vocational pathways.

Vocational education, by contrast, emphasizes technical and applied green skills needed for emerging green industries. Research demonstrates significant variation in vocational curricula across countries, highlighting differences in how green skills are integrated into logistics, agriculture, manufacturing, and other technical fields (Chen, 2025; Hu et al., 2025). Mutohhari et al. (2025), studying Indonesian vocational schools, found uneven levels of green skills maturity among both teachers and students, underscoring challenges in curriculum implementation and professional development. Additional studies show that students’ and teachers’ behavioural engagement in sustainability practices is influenced by institutional support, availability of learning resources, and workplace expectations (Haloho et al., 2023).

Taken together, these findings suggest that green skills constitute a developmental continuum spanning basic, lower secondary, and vocational education. Understanding this continuum requires assessment frameworks capable of capturing differences across educational stages while maintaining conceptual consistency.

2.3 Assessment models for green skills

The assessment of green skills has gained substantial attention as institutions attempt to quantify sustainability competencies for curriculum evaluation and policy planning. Early work in this field focused on environmental behaviour instruments or green character scales, often validated using classical test theory or Rasch modelling (Sukri et al., 2022). These tools emphasized behavioural indicators and attitudinal components, suitable primarily for younger students.

More recent research expands assessment models into vocational and professional contexts. Kamis et al. (2018) developed a psychometrically tested instrument for green skills among Malaysian vocational students, demonstrating acceptable reliability and factorial validity. Meanwhile, Jovanovski et al. (2024) introduced the Green Compass tool for institutional self-assessment, which supports schools in identifying strengths and weaknesses in sustainability implementation. Although useful for organizational evaluation, Green Compass does not measure individual learner competencies and is not applicable for cross-level comparisons.

Sector-specific instruments continue to emerge. Hu et al. (2025) validated green skills for agricultural professionals in China, using SEM-based approaches to confirm construct validity. Similarly, Ibrahim et al. (2020) explored green knowledge and attitudes from an industrial perspective, highlighting the need for alignment between educational preparation and workplace expectations. Research in other applied fields, including green tourism (Ristanović et al., 2025) and informal economic sectors (Manyati et al., 2024), further underscores the importance of robust green skills assessment.

Despite these advances, existing models remain fragmented. Many instruments focus on either early education or vocational contexts, with little attention to cross-level alignment. Furthermore, studies vary widely in their methodological rigor, limiting their comparability.

2.4 Validation approaches in green skills measurement

Instrument validation is critical to ensuring that assessment tools accurately measure intended constructs. Techniques such as exploratory factor analysis (EFA), confirmatory factor analysis (CFA), and structural equation modelling (SEM) have been widely used in educational measurement to assess dimensionality, reliability, and construct validity (Unfried et al., 2015; Lombardi et al., 2022). Within green skills research, these methods have been applied to validate teacher and student competency instruments (Kamis et al., 2018; Mutohhari et al., 2025; Hu et al., 2025). Rasch modelling is also employed to evaluate item difficulty, discrimination, and response patterns (Sukri et al., 2022; Ismail et al., 2024).

However, few studies extend validation to cross-group comparisons. Without robust invariance testing, differences in green skill scores may reflect measurement artifacts rather than true variations across educational levels, genders, or institutional contexts. As sustainability transitions require coherent educational strategies, measurement tools must be validated for comparisons across diverse learner populations.

2.5 Measurement invariance in educational assessments

Measurement invariance ensures that an instrument measures the same construct consistently across different groups. This is essential when comparing competencies across educational levels such as primary, lower secondary, and vocational students. Invariance testing typically proceeds through hierarchical stages—configural, metric, scalar, and strict invariance—each establishing increasing levels of equivalence (Sepasgozar, 2023).

Recent methodological studies highlight the importance of invariance testing in both composite models and latent variable frameworks. For instance, Sepasgozar (2023) applied measurement invariance in composite modelling to examine digital technology assimilation, demonstrating how group-level comparisons depend on establishing structural equivalence. Similarly, Lombardi et al. (2022) emphasize invariance as a prerequisite for meaningful interpretation in educational fidelity assessments, while Unfried et al. (2015) apply invariance approaches to STEM attitude measurement.

Although the importance of measurement invariance is recognized in broader educational research, its application within green skills assessment remains limited. No known studies have examined invariance across the full schooling continuum from basic to vocational education. This gap poses challenges for policymakers and researchers seeking to track green skill development over time or compare implementation efforts across school types.

2.6 Research gap and the need for a unified validation framework

Several gaps emerge from the existing literature. First, research on green skills remains concentrated in sector-specific or vocational contexts, with limited attention to early education where foundational skills develop (Sukri et al., 2022; Ismail et al., 2024). Second, most instruments assess isolated aspects of green skills, lacking an integrative developmental perspective that spans multiple educational stages. Third, cross-group validation remains underexplored, as few studies utilize measurement invariance to ensure comparable constructs across educational levels (Sepasgozar, 2023). Fourth, current frameworks do not provide a comprehensive alignment between foundational, transitional, and technical green skills as conceptualized in recent competency models (Wegenberger & Ponocny, 2025).

Given these limitations, there is a clear need for a unified validation framework capable of assessing green skills implementation across basic, lower-secondary, and vocational education. Such a framework must be psychometrically robust, developmentally sensitive, and theoretically grounded, enabling researchers and policymakers to examine the continuity of green skill development within national education systems. This study aims to address these gaps by constructing and validating a cross-level instrument using measurement invariance to ensure structural equivalence across educational groups.

3.1 Research design

This study adopted a multi-phase quantitative design to develop and validate a unified instrument for assessing green skills implementation across basic, lower-secondary, and vocational education. The validation process followed contemporary psychometric standards emphasizing internal structure, construct validity, and cross-group comparability (Rosario-Hernández et al., 2021; Shultz et al., 2020). In accordance with recommended practices in scale development, the research included item generation, expert review, pilot testing, exploratory factor analysis (EFA), confirmatory factor analysis (CFA), and measurement invariance testing using multiple-group structural equation modelling (Thompson, 2016; de Vries et al., 2025).

Figure 1 illustrates the Unified Green Skills Validation Framework developed in this study to assess green skills implementation across basic, lower-secondary, and vocational education. The framework integrates both measurement and structural components to capture the multidimensional and developmental nature of green skills. At the core of the model are six validated green skills constructs: Environmental Awareness (EA), Sustainable Behavior (SB), Critical Sustainability Thinking (CST), Sustainability Participation (SP), Resource Management Skills (RMS), and Technical Green Practices (TGP). These constructs represent a continuum from foundational cognitive and behavioral competencies to applied and technical sustainability skills. The directional flow depicted in the model reflects the developmental progression of green skills, whereby early awareness and behavior (EA, SB) support higher-order reasoning and participation (CST, SP), which in turn facilitate practical resource management and technical green practices (RMS, TGP). The framework explicitly incorporates cross-level equivalence through multi-group measurement invariance testing, including configural, metric, and partial scalar invariance. This ensures that the constructs are interpreted consistently across educational levels, enabling valid comparisons of latent means. The lower section of the figure visualizes the application of the framework across three educational groups—primary, lower-secondary, and vocational—highlighting differences in dominant green skill profiles at each level. While foundational competencies are emphasized in primary education, vocational education demonstrates stronger technical and applied green skills.

Figure 1. Unified green skills validation framework across educational levels.

3.2 Participants and sampling procedures

A total of N = 215 participants were recruited across three educational levels: basic (primary school), lower-secondary (junior high school), and vocational education. Stratified cluster sampling was used to ensure proportional representation across groups despite differing school sizes. Sample size considerations followed recommendations for CFA and invariance testing, where minimum samples above 200 are considered adequate for model stability and parameter accuracy (van Dijk et al., 2022; Martin, 2013; Owusu et al., 2025). Participation was voluntary and informed consent procedures were applied.

3.3 Instrument development

3.3.1 Item generation

An initial pool of items was generated from literature on green skills, sustainability competencies, vocational education frameworks, and prior validated scales. The goal was to capture the developmental continuum of green skills from foundational to technical competencies. Construction followed accepted principles of clarity, age appropriateness, and construct representation (Shultz et al., 2020; Steffgen et al., 2020). After reviewing theoretical and empirical sources (Kamis et al., 2018; Sukri et al., 2022; Hu et al., 2025), an initial set of 48 items was drafted, distributed across six conceptual constructs:

3.3.2 Content validity

Content validity was established through a panel of 10 experts in green skills, TVET, environmental education, and psychometric assessment. Experts rated each item for relevance, clarity, and representativeness. Items not meeting recommended criteria for content validity were revised or removed following established practices in similar validation studies (Koob et al., 2025; González-Navarro et al., 2023). After this stage, the final pool consisted of 36 items across 6 constructs.

3.4 Pilot testing and exploratory factor analysis

A pilot test was conducted with n = 60 students to evaluate item functioning and preliminary factor structure. EFA using principal axis factoring with oblique rotation was conducted to examine dimensionality, following psychometric recommendations for early-stage scale refinement (Merino-Soto et al., 2022; Avinç & Doğan, 2024). Items with low communalities (<0.40), cross-loadings, or lack of theoretical coherence were discarded. The EFA supported a six-factor solution consistent with the conceptualization of green skills.

3.5 Main study and confirmatory factor analysis

The full sample (N = 215) was used to perform CFA with robust maximum likelihood (MLR) estimation to validate the internal structure of the instrument. Fit indices (CFI, TLI, RMSEA, SRMR) were interpreted according to widely accepted guidelines in psychological measurement (Rosario-Hernández et al., 2021; de Vries et al., 2025). Model adjustments were carefully considered using modification indices guided by theoretical and conceptual consistency. The final CFA model demonstrated satisfactory fit across all indices, confirming the structure of six interrelated green skills constructs.

3.6 Measurement invariance testing

To determine whether the green skills constructs were comparable across educational levels (primary, lower-secondary, vocational), multiple-group CFA was conducted. Following Thompson (2016), invariance was tested hierarchically: (a) Configural Invariance: Assessed whether the baseline six-factor structure was consistent across groups. (b) Metric Invariance: Factor loadings were constrained equal across groups to evaluate equivalence in item–construct relationships (van Dijk et al., 2022). (c) Scalar Invariance: Item intercepts were constrained equal to test whether differences in observed scores reflected true latent differences (Martin, 2013; Rosario-Hernández et al., 2021). Model comparisons used ΔCFI ≤ 0.01 and ΔRMSEA ≤ 0.015 as criteria for acceptable invariance, consistent with recommendations for multi-group assessment in education research (van Dijk et al., 2022). Partial invariance was pursued in cases where full invariance could not be achieved.

3.7 Reliability and construct validity

Reliability was assessed using Cronbach’s alpha and composite reliability (CR), both widely used in recent validation studies (Koob et al., 2025; González-Navarro et al., 2023). Convergent validity was examined through average variance extracted (AVE ≥ 0.50), while discriminant validity was assessed using Fornell–Larcker and HTMT criteria (Shultz et al., 2020). All six constructs demonstrated adequate reliability and validity indicators.

3.8 Additional analyses

Rasch-based item diagnostics were performed to identify potential issues in response patterns and item functioning, following procedures applied in digital literacy and self-regulation assessments (Avinç & Doğan, 2024; Merino-Soto et al., 2022). Exploratory demographic comparisons were also conducted, including gender-based differences, consistent with contemporary research on sustainability communication and competency variation (Ali et al., 2025).

3.9 Ethical considerations

All procedures involving human participants were conducted in accordance with established ethical standards. Ethical approval was secured from the relevant institutional review boards, and research authorization was formally granted by Universitas Negeri Yogyakarta under Letter No. B/3088/UN34.13/TU.12/2025. Prior to data collection, informed consent for participation in the research was obtained from all participants. As the study involved students across primary, lower-secondary, and vocational education levels, the majority of participants were minors. For these participants, written informed consent was obtained from parents or legal guardians, and assent was obtained from the students themselves after a clear explanation of the study objectives, procedures, voluntary nature of participation, and confidentiality assurances. Participants were informed that their participation was entirely voluntary, that they could withdraw at any time without any consequences, and that all data would be anonymized and used solely for research purposes. All consent and assent procedures were administered in written form to ensure ethical compliance, transparency, and proper documentation (andryananda et al., 2025; Fan et al., 2025).

4.1 Preliminary analyses and descriptive statistics

Prior to analysis, the dataset (N = 215) was inspected for missing values, outliers, and normality assumptions. Missing data (<2%) were imputed using the expectation–maximization method. Skewness and kurtosis values fell within acceptable limits, supporting the use of robust maximum likelihood estimation in subsequent CFA, consistent with methodological recommendations in psychometric research (Baştuğ, 2025; Nurumov et al., 2022).

Table 1 presents the descriptive statistics and correlations among the six green skills constructs. As shown in Table 1, correlations ranged from .28 to .58, indicating moderate conceptual relatedness without multicollinearity, which supports discriminant validity and prepares the model for CFA. Moderate correlations indicate each construct is empirically distinguishable while remaining theoretically connected—an expected pattern in multidimensional competency frameworks (Maertens et al., 2024).

4.2 Exploratory factor analysis

Exploratory factor analysis (EFA) on pilot data (n = 60) revealed a stable six-factor solution. As shown in Table 2, all factors showed strong loadings (>.55) and collectively explained 72.4% of variance. These results align with prior studies validating competency-based instruments using multidimensional structures (Galiana et al., 2020).

Factor loadings, eigenvalues, and percentage of variance explained for each green skills dimension during preliminary exploration. The stable factor structure supports the theoretical assumption that green skills develop across foundational, behavioural, and technical dimensions (Maulana et al., 2023; Hjort et al., 2021).

4.3 Confirmatory factor analysis

Confirmatory factor analysis on the full dataset validated the six-factor model. As shown in Table 3, all fit indices exceed recommended benchmarks, indicating excellent model fit.

Goodness-of-fit indices demonstrating the adequacy of the six-factor green skills measurement model. The CFA results validate the six-factor structure, in line with contemporary psychometric instrument validations (Omar et al., 2025; De Beer et al., 2024).

4.4 Reliability and validity evidence

Table 4 presents internal consistency and convergent validity indicators. All constructs demonstrated strong Cronbach’s alpha, composite reliability, and AVE values exceeding recommended thresholds. High reliability and AVE values confirm strong construct integrity, aligning with validation patterns observed in social-emotional and cognitive competency assessments (Furlong et al., 2020).

4.5 Measurement invariance across educational levels

Measurement invariance across the three groups—primary, lower-secondary, vocational—was tested using hierarchical multi-group CFA.

Table 5 presented Summary of configural, metric, scalar, and partial scalar invariance models with their respective fit indices. Partial scalar invariance allows meaningful latent mean comparisons, consistent with findings in developmental and cross-cultural measurement studies (van Dijk et al., 2022; Omar et al., 2025).

4.6 Structural path invariance — Multigroup SEM differences across levels

Table 6 Presented that the model was tested across three groups (Primary, Lower-Secondary, and Vocational). The table reports standardized coefficients (β), robust standard errors (SE), z-values, and p-values for each group, along with an indicator of whether each coefficient is invariant (i.e., not significantly different across groups) based on multi-group SEM comparisons (Wald test/chi-square difference; α = .05). Invariance decisions are determined using p > .05, indicating that the parameter does not significantly differ across groups.

4.7 Latent mean comparisons

Table 7 summarizes latent mean differences across educational levels, using primary-level learners as the reference group (β = 0). The results empirically support the theoretical progression from foundational to technical green skills across educational pathways.

Table 8 show the primary group is used as a reference (latent mean = 0). The table shows the estimated latent means for Lower-Secondary and Vocational, SE, z, p, and Cohen’s d (effect size) for comparison with Primary.

4.8 Additional psychometric analyses

Rasch diagnostics indicated stable item functioning across participants, with INFIT and OUTFIT values within the acceptable range. This aligns with patterns observed in digital literacy and self-regulation instruments validated using Rasch modelling (Avinç & Doğan, 2024; Merino-Soto et al., 2022). No significant gender-based DIF was detected, consistent with recent findings on green communication patterns across demographics (Ali et al., 2025).

5.1 Summary of key findings

The validated six-factor structure provides evidence that green skills encompass multidimensional competencies including environmental awareness, sustainable behaviour, critical sustainability thinking, sustainability participation, resource management skills, and technical green practices. Consistent with prior literature emphasizing the complexity of sustainability competencies (Saari et al., 2024), the results confirm that green skills are not monolithic but instead comprise interrelated yet distinct domains. Measurement invariance results further revealed full configural and metric invariance and partial scalar invariance across primary, lower-secondary, and vocational groups. This indicates that although the factor structure and item loadings operate similarly across levels, some intercept variations reflect expected developmental differences. Such findings mirror patterns in cross-age and cross-context psychometric studies, where partial invariance is common due to developmental diversity (Omar et al., 2025).

5.2 Theoretical implications for green skills development

The results contribute to the theoretical understanding of green skills as a developmental continuum spanning foundational, transitional, and technical competencies. Primary-level learners demonstrated stronger environmental awareness, whereas vocational students showed the highest proficiency in technical green practices. This progression aligns with existing frameworks of environmentally responsible behaviour and sustainability learning trajectories, which propose that awareness and attitudes form early while technical and applied skills emerge later through contextualized learning experiences (Tao, 2025). Furthermore, the increasing strength of constructs such as critical sustainability thinking and participation among older learners supports the notion that cognitive maturity, exposure to real-world problems, and experiential learning environments are central to sustainability-related competence development. This is consistent with broader educational research showing that contextual exposure and guided inquiry foster higher-order competencies (Martinez & Ellis, 2023; Sutrisno et al., 2025).

5.3 Psychometric rigor and contributions to measurement theory

By integrating the principles of modern test theory (McDonald, 2013) with contemporary SEM-based validation procedures, this study reinforces the necessity of robust measurement design for sustainability education instruments. The strong reliability indices (Cronbach’s α and CR) and adequate convergent/discriminant validity demonstrate that the instrument adheres to psychometric standards required for developmental and comparative educational research. Importantly, the application of measurement invariance analysis establishes that observed differences across educational levels reflect true latent differences rather than measurement bias. This significantly enhances the interpretability of green skills comparisons and supports valid inferences, consistent with best practices in cross-group assessment (Rashideh et al., 2025; Anthony Jr et al., 2023).

5.4 Educational implications across school levels

One of the most significant contributions of this study is demonstrating that green skills can be consistently measured across different school levels using a unified framework, something not previously addressed in sustainability education research. The findings highlight several implications:

5.5 Methodological contributions

Methodologically, this study offers a comprehensive approach to validating sustainability-related instruments by combining classical factor analytic techniques with invariance testing. Such approaches align with emerging trends in psychometric scale development where researchers employ multi-method validation to strengthen empirical claims (Rashideh et al., 2025; Martinez & Ellis, 2023). The use of MLR estimation, Rasch diagnostics, and multi-group CFA strengthens the robustness of findings and demonstrates methodological sophistication comparable to recent studies in technology adoption (Anthony Jr et al., 2023) and artificial intelligence–based learning readiness (Sutrisno et al., 2025). Furthermore, the high survey response quality aligns with evidence that well-structured sustainability and competency assessments achieve higher engagement and valid responses when designed with clear constructs and developmental relevance (Holtom et al., 2022).

5.6 Practical and policy implications

The validated unified framework holds practical significance for educators, school administrators, and policymakers.

5.7 Limitations and future directions

Despite its contributions, the study has limitations. First, the sample size, although adequate for CFA and invariance analysis, could be expanded to enhance generalizability across regions. Second, the cross-sectional nature of the data restricts examination of longitudinal growth patterns, which future studies should address to capture developmental trajectories over time. Third, contextual school variables such as teacher expertise or institutional green policies were not incorporated into the model, limiting the explanatory scope regarding sources of variance in green skills.

Future research may expand the framework to include:

6.1 Policy implications

Several implications emerge for policymakers, curriculum designers, and educators:

Overall, the unified framework offers a robust empirical foundation for advancing sustainability education and shaping policies aimed at cultivating environmentally responsible future generations.