A production-grade macroeconomic nowcasting engine designed to estimate official GDP prints weeks before their release. By solving the "ragged edge" problem of mixed-frequency economic data, this engine extracts a real-time, latent "Economic Health" factor using State-Space modeling.
Macroeconomic data is released with significant publication lags. To operate effectively, quantitative strategies require an understanding of the economy's state today, not three months ago. This project utilizes a Mixed-Frequency Dynamic Factor Model (DFM) paired with a Kalman Filter to continuously update GDP estimates as higher-frequency proxies (payrolls, industrial production, retail sales) are published.
The system is built on two primary state-space equations:
- Observation Equation: Maps noisy, mixed-frequency indicators to the latent economic factor.
- Transition Equation: Models the autoregressive evolution of underlying economic health over time.
The Expectation-Maximization (EM) algorithm combined with Kalman smoothing handles missing observations naturally, projecting the latent state forward.
- Data Ingestion:
pandas-datareader, FRED API - Econometrics:
statsmodels.tsa.statespace.dynamic_factor_mq - Data Processing:
numpy,pandas
For Educational and Research Purposes Only. This software is provided "as is" and is not intended as financial, investment, or trading advice. The models and forecasts generated by this repository are for demonstrative purposes of macroeconomic principles and econometric modeling. Do not use these signals for live trading without independent verification and risk management.