What Drives Evidence Uptake in Global Bureaucracies? UNDP’s Acceptance of Evaluation Findings

Over recent decades, evaluation has become a central instrument for producing policy-relevant evidence, enabling accountability, and supporting organizational learning in governance systems around the globe. In International Organizations (IOs), the institutionalization of evaluation since the 1980s has made externally produced evaluation reports a routine source of expert knowledge in administrative decision-making. Yet we still know surprisingly little about what determines whether such knowledge is accepted and translated into organizational commitments once it enters bureaucratic procedures. This paper studies the organizational uptake of evaluative knowledge in the United Nations Development Programme (UNDP) through evaluation management responses: formal, published statements by program staff indicating whether recommendations are accepted and often specifying intended follow-up actions. Conceptually, management responses provide a tractable indicator of organizational acceptance and commitment—a proximal outcome in the governance of evidence that sits between knowledge production and (often untraceable) evaluation use. Empirically, I compile a new dataset from UNDP’s public evaluation repository covering 44,462 recommendation–management-response pairs from 5,965 evaluation reports. The paper measures two key outcomes. First, it codes stated recommendation acceptance, combining structured UNDP labels (available for a subset of recent evaluations) with a validated large-language-model classifier applied to paired recommendation and response texts. Second, it measures response precision, capturing whether management responses contain actionable commitments (e.g., concrete steps, responsibilities, and timelines) rather than generic acknowledgements. To explain variation in acceptance and precision, the analysis integrates predictors that map directly onto debates on evaluation use: (1) project-level features of the evaluated interventions; (2) country-context characteristics shaping incentives and feasibility; (3) evaluation characteristics, including methods used, evaluation budget, and a text-based measure on how positively the intervention is assessed (following Eckhard et al., 2023); and (4) evaluator characteristics (name-linked affiliation and experience across organizations). Statistical models account for the nesting of recommendations within evaluations and country units. This paper advances debates on evaluation use and evidence-based policymaking by introducing organizational acceptance as a measurable intermediate step between evaluative knowledge production and downstream (often unobserved) use. It further leverages validated, large-scale text-based measures with rich project-, evaluation-, and evaluator-level predictors to explain when and why evaluative evidence is accepted in a large international bureaucracy.