Disaster management is a multidisciplinary and multilevel process designed to mitigate the impact of disasters through preparedness, response and recovery strategies. Disaster management involves intricate planning, coordination and governance frameworks that span various levels of government and community stakeholders (Fathani, Putera & Mohamed 2024; Ryan et al. 2020). The Sendai Framework for Disaster Risk Reduction 2015–2030 articulates a perspective that emphasises the importance of robust disaster risk governance, improved understanding of risks and enhancement of resilience among individuals and essential infrastructure. This framework promotes the incorporation of DRR into national and local development plans, especially within the education sector, by identifying schools as essential locations for risk governance and preparedness (Pearson & Pelling 2015).

In this study, DSSG refers to an integrated school governance approach that combines risk knowledge, leadership, stakeholder participation, operational preparedness, monitoring, accountability and institutional learning. The term ‘school resilience’ is used to describe the broader capacity of a school to anticipate, absorb, adapt to and recover from disaster risks, while ‘institutional capacity’ refers to the organisational resources, procedures and accountability mechanisms that enable DSSG implementation.

Conventional frameworks of the disaster management cycle delineate four interconnected phases – mitigation, preparedness, response and recovery – as essential elements of thorough emergency management (Carter 2008). Recent studies challenge the overly linear conception of this cycle, which overly defines the phases and minimises the significance of feedback loops.

Systematic assessments indicate that effective disaster management depends on the integration of tasks within organisational frameworks, decision-making processes and cross-sector coordination, highlighting governance as a critical factor rather than solely as an operational concern (Khan et al. 2023).

This study situates DSSG within a broader transition from a hazard-focused, response-driven model to a risk-governed framework that prioritises proactive capacity, institutional learning and accountability in the educational context.

The risk governance framework highlights that disasters exhibit complexity, uncertainty and ambiguity, necessitating inclusive decision-making processes alongside technocratic solutions (Pearson & Pelling 2015). Also, Renn (2015) posits that stakeholder and public engagement are essential for compelling and legitimate risk governance, recommending participation formats that are adapted to different levels of complexity and uncertainty. Nowell et al. (2018) assert that governance encompasses self-organising and steering processes within complex societies; it is characterised by adaptability, flexibility and openness to cross-sectoral participation.

These have facilitated the emergence of community-based disaster risk management (CBDRM), emphasising local knowledge, social networks and the collaborative development of risk reduction strategies. Systematic reviews indicate that community engagement enhances early warning uptake, evacuation decisions and preparedness behaviours when integrated into ongoing participatory processes, rather than through isolated awareness campaigns (Ryan et al. 2020). There remains ongoing debate regarding the institutionalisation of participation: whether local actors are genuinely empowered to influence priorities or are invited to execute decisions predetermined at higher levels.

Schools hold a critical role in this discourse. They serve as education providers, community hubs and, in numerous countries, designated evacuation centres. School disaster management occupies a critical position at the convergence of risk governance and educational governance. The concept of DSSG emphasises that schools should not only adhere to externally established DRR procedures, but also create their own internal frameworks for leadership, role clarity, stakeholder engagement (including students, teachers, parents and local government) and ongoing self-assessment.

Resilience is a fundamental concept in modern disaster research. Manyena (2006) characterises resilience as the ability of a system – comprising individuals, communities or institutions – to withstand disturbances, adapt and transform while preserving or improving its fundamental functions. This definition critiques limited perspectives of resilience that regard it merely as a return to a pre-disaster condition, emphasising instead the concept of bouncing forward through processes of learning and innovation.

Resilience, viewed through a social-ecological lens, is rooted in intricate interactions among environmental, economic and institutional spheres (Fathani et al. 2024; Folke 2016). In hazardous areas, school resilience encompasses not only the safety of the buildings, but also the ability to operate the school effectively. These include the capacity of school leaders to make prompt decisions amid uncertainty, coordinate with external agencies, manage conflicting demands when schools serve as shelters, and maintain educational continuity through alternative methods.

The literature indicates a tension between employing resilience as a descriptive concept and its translation into operational, measurable dimensions. This tension directly drives the development of assessment tools aimed at capturing the governance aspects of resilience in educational institutions, extending beyond infrastructural safety and basic emergency procedures.

Indonesia, prone to frequent earthquakes, tsunamis, floods and volcanic eruptions, has enhanced its national disaster management system through normative measures. However, empirical evaluations reveal discrepancies between established regulations and their practical implementation. Inan, Beydoun and Pradhan (2018) demonstrate that a formal disaster management cycle is in place, and the results regarding the reduction of loss of life and economic damage are inadequate due to challenges in implementation and insufficient institutional coordination.

In the education sector, various school safety and disaster preparedness programmes are present; however, evidence regarding the extent to which schools have established mature governance capacities – explicit delegation of DRR responsibilities, incorporation of DRR into school development plans and systems for ongoing evaluation – is limited.

This context underscores the importance of assessment tools that effectively address governance quality and institutional arrangements within disaster-prone educational systems, particularly in developing nations. In response to the necessity for systematic evaluation, researchers have created various instruments to assess a school’s capacity to manage disasters.

The School Disaster Resilience Assessment (SDRA) is a prominent tool for assessing the disaster resilience of educational systems. This analysis examines five areas: physical conditions, human resources, institutional issues, external relationships and natural conditions (Shaw et al. 2016).

This emphasises the resources available to schools (such as infrastructure, equipment and documented plans) and their responses during drills or emergencies, rather than the operation of governance processes. Dimensions like leadership commitment and clarity of roles and responsibilities are frequently represented only indirectly or omitted entirely. Frequently, we only document roles and responsibilities, decision-making procedures, accountability mechanisms, stakeholder participation in planning and lessons learned from near-miss events indirectly or not at all. Comparative reviews suggest that numerous tools have been developed within specific hazard or cultural contexts and may require adaptation for broader application in diverse, resource-limited environments.

These indicate a gap in both concept and methodology. Although tools for evaluating school disaster resilience are available, there remains a scarcity of assessment frameworks that are explicitly based on risk governance theory and specifically designed for assessing DSSG. Few tools effectively integrate principles from risk governance, such as inclusiveness, transparency and responsiveness, alongside resilience thinking, which includes learning, adaptability and transformability and educational management. This integration is crucial for developing a comprehensive set of governance indicators that can assess the maturity of disaster management practices at the school level and inform continuous improvement.

The literature on disaster management suggests that technical capacity alone is insufficient; governance structures, community engagement and institutional learning have a profound impact on outcomes. In the realm of DRR in school, particularly in developing nations like Indonesia, there exists a deficiency of assessment tools that is frame school as proactive governance entities within comprehensive disaster management frameworks, rather than as mere recipients of national directives; evaluate governance quality and institutional maturity, in addition to physical safety or preparedness initiatives; and flexible to various hazard profiles and resource contexts, while maintaining conceptual integrity and measurement reliability.

To address these gaps, it is essential to develop an assessment framework that focuses explicitly on DSSG, converting theoretical principles of risk governance and resilience into practical, context-sensitive indicators. This framework can assist schools, local governments and national policymakers in evaluating governance performance, pinpointing key areas for capacity enhancement and establishing ongoing improvement in education systems vulnerable to disasters.

Based on risk governance, resilience governance and the CSSF, this study derives five governance dimensions that are expected to be associated with DSSG: Leadership and Strategic Commitment (X1), Risk Knowledge (X2), Stakeholder Participation (X3), Operational Preparedness/Resource Management (X4) and Learning, Monitoring and Accountability (X5). These dimensions represent the institutional, cognitive, participatory, operational and evaluative foundations of school-based DRR. Accordingly, the following hypotheses are proposed:

Ethical clearance to conduct this study was obtained from the Universitas Andalas Research Ethics Committee (Ref. No. B/468/UN16.08.D/TP.02.02/2025).

The study comprised 215 participants, including principals, deputy principals, teachers, administrative personnel and school committee administrators. Table 2 presents the proportion of respondents, which closely correlates with the proposed variables and indicators. The respondents’ findings suggest that they are currently in a productive phase and have adequate experience in school governance, particularly in implementing governance and policies for disaster-safe education.

The PLS algorithm assessed the validity and reliability of the research data, demonstrating consistency with the statements and categorising it as a valid instrument that meets the practical measurement requirements of the research variables. The external load, when assessed against the binding indicators, was determined to be valid and acceptable, with a value exceeding 0.60.

The indicators utilised demonstrated the reliability of the variables and the acceptability of the measurement model. The acceptance of indicators suggests the absence of conflicts that could undermine other pertinent coefficients (refer to Table 3).

Indicators X1.5 and X4.5 were removed because their outer loading values did not meet the acceptable measurement threshold and weakened the overall measurement quality of their respective constructs. The removal was conducted to improve indicator reliability, convergent validity and the parsimony of the measurement model, while retaining the conceptual integrity of leadership and strategic commitment (X1) and operational preparedness/resource management (X4). The results of the convergent validity test indicated that nearly all indicators exhibited loading factor values exceeding 0.60, specifically ranging from 0.671 to 0.935. Additionally, the AVE values for each variable fell within the range of 0.570 to 0.832, thereby satisfying the established measurement threshold.

Additionally, internal reliability was assessed through Cronbach’s alpha and composite reliability (CR), with a minimum threshold of 0.70. The results for all items met the threshold, with Cronbach’s alpha values ranging from 0.861 to 0.935 and composite reliability values ranging from 0.853 to 0.940. The research instrument is deemed reliable, confirming that the indicators employed effectively elucidate the latent variables.

To provide a more complete reporting of the PLS-SEM measurement model, convergent validity, internal consistency reliability and discriminant validity were assessed before interpreting the structural model. Convergent validity was confirmed through outer loading values above the acceptable threshold and AVE values exceeding 0.50. Internal consistency reliability was supported by Cronbach’s alpha and composite reliability values above the recommended threshold. Discriminant validity was also examined to ensure that each latent construct was empirically distinct from the others.

Each dimension possesses distinct criteria and rationale for mutually reinforcing one another within the research instrument. Each indicator employed exerts a substantial influence on the measurement. The indicators for variable X1 exhibit a value range of 0.779 to 0.882, with indicator X1.2 being the predominant one, as it effectively elucidates and represents the commitment of the dominant leadership. Variable X2, ranging from 0.844 to 0.930, is primarily influenced by indicator X2.3, which highlights the importance of two-way communication and rejects hierarchical or top-down methodologies.

Variable X3, exhibiting a value range of 0.671 to 0.848, is primarily influenced by indicator X3.2, which underscores the necessity of full participation from all students in achieving effective disaster management outcomes. Variable X4, exhibiting a value range of 0.683, is predominantly influenced by indicator X4.2, which highlights the routine simulation practices implemented by the local government and mandates participation from all students. Variable X5, which ranges from 0.728 to 0.902, is primarily influenced by indicator X5.1, highlighting the importance of routine evaluation and monitoring of all policies approved by stakeholders. The external content demonstrates a mutually reinforcing correlation, indicating that risk enhancement is essential for developing a DSSG assessment.

In PLS-SEM, the inner model in direct effect analysis indicates the quality, accuracy, strength and feasibility of the structural model that delineates the relationships among latent constructs. Consequently, it necessitates a set of stringent statistical and substantive evaluation criteria. The fulfilment of the VIF, R², p-values for significance of path coefficients, F², Q² and standardised root mean square residual criteria collectively underpins the argumentation of the estimated inner model, demonstrating its adequate explanatory, predictive and stability capabilities to substantiate the theoretical findings and empirical implications of the research.

The structural model should be free from collinearity issues by ensuring that the VIF values for the exogenous constructs remain below the recommended threshold of 5. These ensure that the path coefficient estimates are accurate and that collinearity does not compromise these estimates (Hair et al. 2019). The findings indicated that all exogenous constructs displayed VIF values below 5, with a range from 1.339 to 4.637. These suggest the absence of multicollinearity issues, thereby confirming the reliability of the regression coefficient estimates.

The structural model was evaluated through collinearity assessment, path coefficient significance, explanatory power, effect size and predictive relevance. Collinearity was assessed using VIF, while the significance of structural paths was examined through bootstrapping. R² was used to assess explanatory power, F² to evaluate the relative contribution of each exogenous construct, and Q² to examine predictive relevance. These procedures were included to provide a more transparent reporting of the PLS-SEM analysis.

The relationships between constructs were evaluated using a bootstrapping procedure, focusing on the t-statistic and p-value of the path coefficients. The relationships between constructs were evaluated using a bootstrapping procedure, focusing on the t-statistic and p-value of the path coefficients. Statistically significant and directional coefficients provide empirical support for the hypothesised structural relationships within the research model (Hair et al. 2019; Henseler & Sarstedt 2013). However, given the cross-sectional nature of the data, the results are interpreted as evidence of association among constructs rather than as definitive structural association. The constructs’ strength was evaluated using the coefficient of determination (R²), categorised as substantial (R² > 0.75), moderate (0.50 < R² < 0.75) and weak (R² < 0.25) (Hair et al. 2017; Sarstedt et al. 2019). The F² effect size was employed to evaluate the contribution of each exogenous construct to the endogenous construct in the direct structural relationship (Cohen 1988; Hair et al. 2019). Refer to Table 4 for the results of the structural evaluation.

The findings indicated that X1 exerted a positive effect of 17.2% (0.172), which was statistically significant with a p-value of 0.001. X1 exhibited a minimal effect on Y, indicated by an F² value of 0.061. The confidence interval (CI) extended from 0.062 to 0.270, indicating a significant relationship at the 95% confidence level. These findings suggest that robust strategic commitment and leadership patterns can harmonise policies and effectively integrate them into school governance and disaster management planning.

The analysis of X2 indicated a positive effect of 47.2% (0.472), which was statistically significant for Y, with a p-value of 0.000, falling below the 0.05 threshold. X2 exhibited a moderate effect, indicated by an F² value of 0.289, with a CI ranging from 0.342 to 0.598, which does not encompass zero. The significant value produced is closely linked to the educational practices and routine implementations by stakeholders, particularly given the research area’s susceptibility to various disasters. Education, outreach and both vertical and horizontal relationships should continue to engage the community. X2 demonstrates that this relationship is significant at the 95% confidence level.

The analysis of X3 indicated a similarity to X1, maintaining significance with a p-value of 0.000. The effect size was weak, as indicated by an F² value of 0.056, and a positive correlation of 11.5% (0.115) was observed. This positive correlation stems from community knowledge and local empowerment, leading to coordination, albeit on a limited scale. The influence of X3 exhibited a CI that approached zero, spanning from 0.051 to 0.180. This relationship was significant at the 95% confidence level.

The results of X4 indicated an insignificant outcome, as evidenced by the p-value (0.685) and a negligible effect size, reflected in the F² value (0.001). This is additionally evidenced by a minimal positive correlation of 0.017, or 1.7%, alongside a CI that includes zero, spanning from −0.070 to 0.091. This is evident when a minimal and inconsequential influence is reflected in actual infrastructure practices. It affirms that physical factors alone do not entirely dictate the effectiveness of governance.

X5 demonstrates a positive effect of 27.7% (0.277), which is statistically significant with a p-value of 0.000, indicating it is below the 0.05 threshold. The influence of X5 is moderate, indicated by an F² value of 0.184 and a CI ranging from 0.178 to 0. The relationship is statistically significant at the 95% confidence level. Disaster education and a culture of resilience play a significant role in governance, albeit with a lesser impact compared to governance and capacity.

Variable X4 is the sole construct that is statistically insignificant, as the confidence interval includes zero, and the p-value is above the established threshold. Additionally, constructs X1, X2, X3 and X5 exhibit positive and significant effects, as indicated by the confidence interval not encompassing zero and the p-value being below 0.05.

The appropriateness and assessment of the PLS-SEM test are discussed to determine the model’s predictive capability. The components comprise model measurement statements, such as R², Q², F², PLS Predict and GoF (Hair et al. 2019; Sarstedt et al. 2019). R² represents the coefficient of determination, indicating the proportion of variance in the construct that can be accounted for by the latent constructs within the model (Wetzels, Odekerken-Schröder & Van Oppen 2009). A higher R² indicates a stronger explanatory power for the endogenous construct. The values are classified as 0.75 (substantial), 0.50 (moderate) or 0.25 (weak), contingent upon the context (Hair et al. 2017, 2019).

Q² serves as an indicator of predictive relevance, reflecting the model’s capacity to forecast the indicator values of the endogenous construct (Hair et al. 2014, 2019). The analysis of Q² via blindfolding, under the condition that Q² > 0, demonstrates that the model possesses predictive relevance to the construct. Goodness-of-fit refers to the extent to which a model aligns with the data. Standardised root mean square residual serves as an index for assessing model fit, where a value below 0.08 is deemed fit or acceptable (Henseler & Sarstedt 2013).

The R² value of 0.851 suggests that 85.1% of the variation in variable Y is accounted for by the latent constructs within the structural model. Additionally, only 14.6% of the variation is attributed to external factors not accounted for in the model. This is classified as possessing significant explanatory power. It is essential to situate it within a comprehensive structural evaluation framework that includes significance and path effects, predictive relevance and potential risks, ensuring that claims extend beyond R² figures to align with current debates and best practices in PLS-SEM analysis.

The Q² value demonstrates a significant strength of the observed variables, with a quantified value of 0.846. The robust F² value has a significant influence on all measurement metrics within the research model. The provisions of Henseler and Sarstedt (2013) are valid, as all observed constructs exhibit large dominant effect sizes and meet the established requirements.

The analysis of 22 research indicators across 5 constructs and 20 components, utilising RMSE and MAE, indicated that the measurement items were lower than those produced by the linear regression model. The findings suggest that the proposed PLS model exhibits moderate predictive capability. The research model’s GoF value was 0.249, indicating a moderate fit. Figure 1 presents the research’s structural model.

The final DSSG framework should be read as an integrated evaluation tool for schools and policymakers. Risk Knowledge (X2) and Learning, Monitoring and Accountability (X5) serve as the core dimensions because they show the strongest associations with DSSG. Leadership and Strategic Commitment (X1) and Stakeholder Participation (X3) function as enabling dimensions that support implementation and legitimacy. Operational Preparedness/Resource Management (X4), although not statistically significant in this model, remains relevant as an operational component when embedded within risk-informed planning, accountability and continuous improvement. Therefore, the framework can be used to diagnose governance gaps, prioritise capacity-building interventions and guide school-level DRR planning.

Practically, the DSSG framework can be used by schools and local authorities as an evaluation tool to diagnose governance gaps and prioritise improvement. Schools may use the framework to assess whether risk knowledge is integrated into planning, whether preparedness activities are linked to accountability mechanisms, and whether monitoring results are used for continuous improvement. Local education authorities can also use the framework to identify schools requiring stronger technical support, resource allocation and policy supervision.