Overview

---

Most recession models are reactive, not predictive. They rely on lagging indicators like GDP or unemployment, single-variable triggers like yield curve inversion, or survey data that reflect consensus rather than signal. These models often miss early warnings, fail to adapt to nonlinear shifts, and can’t integrate real-time sentiment or policy guidance.

In today’s macro environment, characterised by faster monetary cycles, geopolitical shocks, and narrative-driven volatility,such models are no longer sufficient.

MACRA.AI is a dynamic, multimodal forecasting architecture purpose-built for modern macro risk. It combines:

• Bayesian probabilistic reasoning
   
• High-dimensional macroeconomic VAR modelling
   
• Deep learning overlays (BiLSTM and hybrid CNN‑LSTM)
   
• Real-time unstructured data ingestion (LLM narrative models)
   
• Scenario-conditioned simulations based on Fed, fiscal, and global inputs
   

This isn’t a tweak to legacy models, it’s a structural overhaul. MACRA.AI is engineered to detect inflection points, quantify uncertainty, and adapt to new data with audit-ready transparency.

The MACRA system integrates five key signal engines:

1. VAR Macroeconomic Blocks – tracking monetary policy shifts, domestic fragility, credit stress, global spillovers, and sentiment divergence. Each block is driven by PCA-compressed macro features and calibrated on post-2000 cycle dynamics.
    
2. Bi‑LSTM Forecast Overlay – a deep learning module that captures regime shifts and nonlinear path dependencies, especially in labor softening, credit deterioration, and inflation transitions.
    
3. LLM Narrative Analytics – transformer-based embeddings (e.g., SBERT, FinBERT) turn unstructured inputs like Beige Book text, FOMC speeches, and corporate earnings sentiment into real-time macro signals. These feed directly into the narrative VAR block and recession priors.
    
4. Scenario Simulator – forward macro paths are generated under multiple conditional branches: 0, 1, or 2 rate cuts; QT acceleration; tariff shock; consumer credit tightening; or geopolitically driven supply disruption. Each branch yields adjusted posterior probabilities.
    
5. Recession Classification Engine – classifies risk across three dimensions:
    
   • Technical recession: two quarters of negative real GDP
      
   • NBER-defined recession: broad-based activity contraction across employment, income, and industrial production
      
   • Felt recession: a composite index using sentiment, hardship, and consumption proxies to reflect public/business perception
      

MACRA.AI has demonstrated superior predictive power across multiple recession windows. It consistently identifies turning points before consensus models. Compared to traditional tools, like the NY Fed's yield curve model or the Sahm Rule, MACRA.AI shows materially lower false positives, better calibration, and higher predictive accuracy, especially in policy-sensitive regimes.

In recent backtests, MACRA outperformed benchmark models across ROC AUC, false positive suppression, and forecast reliability. Its performance is particularly strong in late-cycle conditions where labor and credit signals begin to fragment.

MACRA.AI is designed for real-world institutional decision-making:

• Central banks and regulators use it for monetary transmission lag detection, policy scenario testing, and financial stability planning.
   
• Asset managers and hedge funds use MACRA to map risk-on/off positioning to expected macro inflections and narrative regime shifts.
   
• Banks and credit professionals rely on MACRA for early warning of credit cascade risks, default clustering, and liquidity tightening.
   
• Corporate strategy teams use MACRA for scenario-informed demand forecasting and investment horizon planning.
   

The system also outputs a structured, time-stamped audit trail of each model update, aligned with regulatory model governance standards.

MACRA.AI is designed to meet and exceed supervisory expectations, including:

• U.S. SR 11‑7 (Model Risk Management)
   
• ECB TRIM (Targeted Review of Internal Models)
   
• UK PRA SS1/23 (Supervisory Statement on Model Risk)
   

All model versions are checkpointed, priors are traceable, and input-output mappings are preserved. Scenario conditioning and narrative inputs are recorded and auditable.

MACRA.AI offers a radically modern solution to forecasting macroeconomic risk. It is:

• Mechanically grounded in theory and empirical performance
   
• Adaptive to fast-changing macro, credit, and policy conditions
   
• Transparent and auditable, meeting institutional risk standards
   
• Forward-looking, offering conditional simulation and deep signal integration
   

In a world where economic cycles shift faster and confidence breaks earlier, MACRA.AI provides clarity, foresight, and defensible insight.