AI in central banking: Synthetic vs real world data in forecasting models

Most traditional econometric models used by central banks cannot forecast accurately and on time in any era of major structural changes and increased volatility of economic scenario planning today. In this regard, this study proposes a dual-pipeline design that allows the integration of artificial intelligence (AI) with forward-looking exploratory synthetic data alongside real-world data as an empirical base. We introduce a “Preparedness Maturity Model” to guide this transition, culminating in a “High Preparedness” stage characterized by hybrid models. This includes pre-training deep learning architectures on diverse synthetic scenarios and fine-tuning them on curated historical data. Ultimately, we conclude that the strategic adoption of a hybrid real-synthetic data approach is essential for enhancing the resilience and foresight of monetary policy.