Purpose This paper aims to propose an integrated, systemic and multilevel model of public-sector innovation capability. It maps the construct’s elements at micro (individual), meso (team/organization) and macro (ecosystem) levels, emphasizing the interplay of internal and external factors, and frames capability as a moderating variable between antecedents, the innovation process and expected results. Design/methodology/approach A PRISMA systematic review of Scopus and Web of Science (June 2023–December 2024) retrieved 248 records; 27 met the eligibility criteria. Two independent reviewers conducted thematic content analysis with double coding and reliability checks, synthesizing evidence into a multidimensional framework. Findings Four dimensions and 14 elements emerged: behavioral (human capital, leadership, culture, communication), organizational support (structure, processes, resources, technology), governance (strategy, regulation, risk management) and relational (learning, collaboration, stakeholders). Each element can hinder or catalyze innovation, depending on the institutional context, thereby moderating both antecedent-to-process and process-to-results links. Research limitations/implications Empirical validation and operational instruments are still needed to track the longitudinal evolution of innovation capability. Future research should test the model in varied contexts, develop scales and examine its moderating interactions over time. Practical implications The framework guides managers in diagnosing bottlenecks, aligning strategy, resources and governance and tackling bureaucratic rigidity, silos and asset obsolescence to foster continuous innovation. Social implications Applying the model can lead to more responsive public services, greater public value and enhanced social trust in institutions. Originality/value This is the first model that positions innovation capability as a moderating mechanism, highlights the team level’s critical role and acknowledges the construct’s functional ambiguity.
Soares et al. (Sun,) studied this question.