Clustering-Conditioned Granger Causality Between GDP Growth and Private Financing

Whether finance leads growth or growth leads finance remains a century-long debate. We argue that the direction and strength of the GDP–credit nexus are context-dependent and can be systematically uncovered by conditioning causal analysis on macro-structural heterogeneity across countries. We implement a two-stage pipeline: (i) unsupervised clustering of 30 economies (2005–2022, five annual macro indicators) via Agglomerative Clustering to form homogeneous macro-structural groups; and (ii) within-cluster dynamic causal analysis using lagged correlations, Granger causality and explanatory models (quarterly GDP and private credit, year-on-year growth, 1Q2005–3Q2024). Results show non-universality of causality: (a) in “developed and in transition, economically stable” economies, credit → GDP is predominant; (b) in “highly developed, competitive and stable” economies, bidirectionality is predominant; however, the results are not economically intuitive; (c) in “emerging/intermediate with macro risk”, bidirectional links are common, and feedback between both variables is interpretable across distinct scenarios. Post hoc Lasso and XGBoost confirm effect magnitudes and non-linear thresholds. We contribute a macro-segmented causal discovery framework that reconciles conflicting findings in the literature and provides policy-relevant differentiation by economic context.