BACKGROUND: This study examines the relationships among leadership styles (transformational and paternalistic) and institutional autonomy, cross-sector collaboration, governance system flexibility, and decentralization within the Chinese public sector. While grounded in China's public administration, the findings contribute to broader governance theories by highlighting global mechanisms that may apply to other hierarchical and centralized administrative systems. METHOD: We employed a convenience sampling approach using a sample of 1358 public employees from various government departments across major Chinese municipalities, ensuring representation across diverse regions and departments. Data were collected through a web-based survey and analyzed using descriptive statistics, correlational analyses, and regression models to test four hypotheses. Regression analysis was incorporated to establish the strength and direction of relationships between variables, strengthening the validity of the findings. FINDINGS: Results revealed that institutional autonomy and cross-sector collaboration were positively associated with governance system flexibility and decentralization, supporting Hypotheses 1 and 2. Transformational and paternalistic leadership partially mediated these relationships, supporting Hypotheses 3 and 4. The regression results reinforced these associations, demonstrating that leadership styles are significant but partial mediators between institutional structures and governance outcomes. CONCLUSIONS: These findings underscore the importance of autonomy, collaboration, and leadership styles in fostering adaptable and decentralized governance systems in the public sector. Beyond the Chinese context, these insights may inform governance reforms in other centralized administrative settings, such as those found in transitional economies or large-scale bureaucratic institutions. The study's implications, limitations, and suggestions for future research are discussed, with recommendations for cross-national studies to further validate and expand on these findings.