How predictable are public procurement awards? Network evidence from EU above-threshold contracts

Purpose This study aims to examine the degree to which contract awards in EU above-threshold procurement are predictable from prior buyer–supplier relationships, and whether this predictability varies across procurement domains. Design/methodology/approach Using longitudinal award data from Tenders Electronic Daily (TED, 2018–2022), this study models procurement as an evolving bipartite network across twelve high-activity Common Procurement Vocabulary (CPV) categories. Three network-based features – Preferential Attachment, an adapted Adamic–Adar index and Historical Frequency – capture distinct dimensions of relational structure. These features are used in a supervised classification framework (XGBoost) to predict which authority–supplier pairs will transact in subsequent years, with robustness assessed across three class balance conditions. Findings Predictive performance is consistently strong (mean AUC > 0.96, mean precision > 0.90), but substantial cross-domain heterogeneity emerges. Medical and pharmaceutical categories exhibit F1 scores above 0.75, indicating strong relational persistence. Construction and transport services display lower recall, suggesting more competitive dynamics. Direct relationship history becomes the dominant predictor when determining which supplier is most likely to receive the award. Research limitations/implications The predictive model relies exclusively on network structure and relationship history, without incorporating contract-level attributes such as individual contract value or technical specifications. This partly explains lower predictive coverage in project-driven domains. Practical implications Predictability metrics can support category-specific monitoring, inform supplier diversification strategies and identify where SME access policies require interventions beyond procedural openness. Social implications By revealing where relational persistence limits competitive openness, the findings inform policies aimed at broadening market access for small and medium-sized enterprises and new entrants. In high-predictability procurement categories, interventions beyond procedural openness – such as supplier diversification strategies and targeted monitoring – may be necessary to ensure fairer distribution of public contract opportunities. Originality/value The paper provides the first large-scale, cross-domain measurement of award predictability in EU procurement, demonstrating that link prediction techniques adapted to bipartite procurement networks serve as scalable diagnostic tools for assessing effective contestability.