A component model and indicators of supervisory leadership for primary school administrators in Northeastern Thailand

A mixed-methods study of primary school administrators and academic teachers in Northeastern Thailand identified a five-component configuration for supervisory leadership that fitted the observed covariance structure well. The model organizes supervisory leadership into curriculum and instructional management, positive communication, decision-making ability, supervision evaluation and assessment, and teacher professional development. In a sample of 680 respondents from 340 schools, the final second-order confirmatory factor model produced a comparative fit index and Tucker–Lewis index of 0.999, a root mean square error of approximation of 0.016 and a standardized root mean square residual of 0.014. These results support the proposed structure for this regional sample, but they do not establish a universal model of effective supervision or demonstrate that any component improves teaching or student achievement.

The research used an exploratory sequential mixed-methods design in which a qualitative phase built the instrument tested in a quantitative phase. The qualitative work began with a narrative synthesis of more than 50 domestic and international sources on leadership, supervision, school administration, curriculum and instructional management, evaluation and teacher professional development. The researchers then interviewed nine purposively selected experts in three groups: educational administration scholars, supervision specialists and primary-school practitioners. The interviews confirmed five components and 15 indicators, refining some wording and operational definitions without adding or removing categories. One researcher conducted the coding, while dissertation advisers reviewed the interview guide and preliminary framework, notes and audio records provided an audit trail, and all nine experts took part in a final member-checking exercise. The confirmed components and indicators were converted directly into the questionnaire used in the second phase, with instrument development serving as the integration point between the qualitative and quantitative strands.

The quantitative population comprised administrators and academic teachers in primary schools under the Office of the Basic Education Commission in Northeastern Thailand during the 2025 academic year, a setting encompassing 12,126 schools. Using multistage random sampling, the researchers selected 340 schools across all 20 provinces in the region, allocating quotas proportionally by province and randomly choosing schools within provincial strata. Two designated respondents were selected at each school: its director and one academic-affairs teacher. Data collection took place in February 2026 through electronic questionnaires, producing 680 complete responses with no exclusions or missing data. The instrument contained 44 items scored on a five-point Likert scale from strong disagreement or the lowest level of agreement to strong agreement or the highest. Nine experts assessed its content validity, yielding item content-validity coefficients from 0.73 to 1.00, while a pilot with 30 participants produced a Cronbach’s alpha of 0.93. Scores reflect respondents’ perceived agreement with described supervisory behaviors, not directly observed actions or objectively calibrated levels of competence.

To test the five-part structure through second-order confirmatory factor analysis, the researchers combined the 44 questionnaire items into 15 approximately balanced parcels, allocating three to each component. Sequential assignment distributed items across positions within each component so that each parcel represented a mixture of content. Positive communication and curriculum and instructional management each used nine source items; decision-making ability and supervision evaluation and assessment each used eight; teacher professional development used ten. Parceling reduced the number of first-order indicators to a three-indicator structure for every component and was intended to limit the influence of individual items whose response distributions were non-normal. It also restricted the conclusions: the analysis concerned relationships among parcels rather than individual items, potentially obscured item-level multidimensionality, and could make fit appear stronger than it would in an item-level solution. The researchers did not conduct an item-level robust sensitivity analysis.

The component loadings on the overarching supervisory-leadership factor were high. Curriculum and instructional management had the strongest standardized loading, β = 0.946, with indicators for curriculum understanding, curriculum design and evaluation of curriculum implementation. Positive communication followed at β = 0.918 and comprised creating a positive climate, building shared understanding and providing support. Decision-making ability loaded at β = 0.913, represented by decisiveness, adaptability to change and sound judgment. Supervision evaluation and assessment had a loading of β = 0.904, covering use of diverse supervisory tools, evaluation of teaching outcomes and monitoring teacher development. Teacher professional development loaded at β = 0.884, with indicators for developing teachers’ knowledge and skills, evaluating and reflecting on learning, and promoting exchange of knowledge among teachers. These coefficients describe each component’s statistical relationship with the overall factor in the specified model; they do not measure causal influence, practical priority or the extent to which any behavior directly produces better educational outcomes.

The initially specified model, before residual covariances were introduced, already fitted acceptably. It produced χ² = 106.242 with 85 degrees of freedom, p = 0.0593, RMSEA = 0.019, CFI = 0.998, TLI = 0.997 and SRMR = 0.028. Modification-index-guided refinement added three residual covariances and yielded the final χ² of 95.693 with 82 degrees of freedom and p = 0.143, alongside the higher CFI and TLI of 0.999, RMSEA of 0.016 and SRMR of 0.014. Thus, the changes improved fit only marginally rather than creating it from an initially poor model. Two freed covariances had stated statistical and theoretical rationales: one linked curriculum parcels sharing documentation-related content, and another linked teacher-development parcels involving structured skills development and peer knowledge exchange. The third connected two positive-communication parcels but was not statistically significant and lacked a distinct theoretical justification. The researchers say it should be removed or reconsidered in future reanalysis, and they did not test whether its removal materially changed fit.

Validity evidence was encouraging but not unambiguous. Average variance extracted values ranged from 0.935 to 0.943, and their square roots exceeded the corresponding intercomponent correlations in the Fornell–Larcker comparison. In the heterotrait–monotrait analysis, however, positive communication, decision-making ability and curriculum and instructional management each reached a maximum ratio of 0.869. That was above the conservative 0.85 threshold, although below the more liberal 0.90 threshold, indicating that the separation among those dimensions requires caution. Response patterns also limited the analysis. Across 44 items and 29,920 item-level responses, 69.3% of ratings were 5, 26.7% were 4 and 3.9% were 3; only 0.1% were 2, and no respondent selected 1. Individual items had maximum-category response rates ranging from 52.5% to 89.4%. Thirty-four of the 44 items had skewness magnitudes greater than 1, and a multivariate normality test rejected the normality assumption used by standard maximum-likelihood estimation. The concentrated responses and just-identified three-indicator factors help qualify the exceptionally strong fit indices.

The nested sampling design introduced another uncertainty. Because two respondents came from each of 340 schools, responses from the same school could be correlated through shared perceptions of local supervision, yet the analysis did not use cluster-adjusted standard errors or a design-effect check. The retained dataset lacked the school identifier needed to link paired respondents for retrospective correction, so the reported precision and fit statistics may not fully reflect the sampling design. Administrator and teacher responses also showed different patterns. Teachers rated curriculum and instructional management, supervision evaluation and assessment and teacher professional development significantly higher than administrators, with all p < 0.001, while positive communication and decision-making ability did not differ significantly between the groups. Intercomponent correlations ranged from 0.22 to 0.80 among administrators but from 0.75 to 0.94 among teachers, a pattern the researchers said might indicate a stronger halo tendency in teacher ratings. No formal configural, metric or scalar invariance analysis was performed, so the pooled model cannot be assumed to measure the construct equivalently for self-raters and teacher-raters.

Within these limits, the study offers a context-specific measurement framework for supervisory leadership in Northeastern Thai primary schools. Its strongest statistical components connect the academic core of supervision with trust-based communication, judgment, evaluation and support for teachers’ continued learning, while the mixed-method design provides both content-development evidence and a test of internal structure. The cross-sectional data do not establish whether stronger perceptions of any component lead to better teaching, teacher performance or student learning, and the single-region sample cannot establish transferability to other parts of Thailand or other countries. The evidence instead supports a promising empirical configuration for further testing. The researchers identify several unfinished analyses: re-estimation with a non-normality-corrected estimator such as MLR, cluster-adjusted inference, a formal common-latent-factor comparison and multi-group invariance testing across administrators and teachers. They also call for outcome-linked and quasi-experimental studies of professional-development programs and replication in other education systems. Until those steps are completed, the five-component model is best treated as a regionally supported measurement proposal rather than a settled standard for judging school leaders.

Subject of Research: Empirical development and testing of a five-component supervisory-leadership model among primary school administrators and teachers in Northeastern Thailand.

Article Title: A component model and indicators of supervisory leadership for primary school administrators in Northeastern Thailand

Article References: Ngamying, W., Sirisooksilp, S., & Ariratana, W. (2026). A component model and indicators of supervisory leadership for primary school administrators in Northeastern Thailand. Discover Education, 5(1), Article 1203. https://doi.org/10.1007/s44217-026-02270-7

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02270-7

Keywords: supervisory leadership, primary schools, Northeastern Thailand, school administration, instructional supervision, teacher development, curriculum management, positive communication, confirmatory factor analysis, mixed methods, component, model