Monetary Dynamics and Inflation Persistence: A Machine Learning Approach Applied to Turkey

This study re-assesses the root causes of inflation through an innovative hybrid analytical framework integrating deep neural networks, random forest algorithms, and causal inference (DoWhy) within the Quantity-Theoretical Inflation Theory, using monthly data from Turkey for the period 2003M05–2025M01. The findings reveal that inflation expectations and exchange rate fluctuations are the most influential drivers of inflation, overshadowing other variables such as money supply, real interest rates, and global oil prices. Although the exact ordering of predictors is model dependent, and the random forest assigns the leading role to the real interest rate, the signal shared across all three methodological pathways rests on expectations and the exchange rate. While money supply growth aligns with monetarist theory and remains a consistent source of upward price pressure, its impact is often mediated through expectations and currency depreciation. Causal estimates confirm that Turkey’s persistent inflation is fueled not just by macroeconomic imbalances but by the erosion of policy credibility and weak anchoring of expectations, particularly after 2017. The study contributes methodologically by introducing AI-powered modeling to inflation analysis in emerging markets, and empirically by validating the central role of expectations and exchange rate pass-through in a structurally fragile, import-dependent economy. The results suggest that sustainable price stability in Turkey hinges on reestablishing central bank independence, adopting orthodox inflation targeting, and regaining public trust through consistent and transparent communication.