Basel Point — Paper 3: Supervised Logistic Model with Macro Variables

This repository provides the public companion materials for Basel Point — Paper 3:

“Why the Yield Curve Lost Its Signal — And How to Recover It”
Substack: https://thebaselpoint.substack.com/p/why-the-yield-curve-lost-its-signal

It includes a redacted preview notebook and static figures illustrating the model’s key outputs.
The full modeling pipeline (data engineering, full notebook, internal diagnostics) is maintained in a private research repository.

---

Repository Contents

Redacted Notebook paper3_macro_supervise_redacted.ipynb
A preview notebook showing:

• feature set overview (yield spreads + macro indicators)
• model structure (12-month-ahead recession probability using logistic regression)
• selected figures
• explanatory commentary

All sensitive components (data loading, feature engineering code, validation loops, and scenario procedures) have been intentionally removed.

---

Figures Static PNGs used in the Substack article:

• ROC_curves_macro_extended_models.png
• scenario-based_conditional_recession_probabilities.png

These illustrate the predictive lift from macro-supervised models and the recession probability bands under different macro-economic scenarios.

---

Method (High-Level Overview)

This paper extends yield-curve recession modeling by incorporating macro-supervised features, including:

• Yield curve signals

  • 10Y–3M spread
  • 10Y–2Y spread

• Macro indicators

  • ISM-style business cycle measures
  • Labor market dynamics (initial claims, unemployment deltas)
  • Credit spreads (BBB, TED)
  • Stress & volatility proxies (e.g., MOVE, VIX)

• Target variable

  • NBER recession indicator, shifted 12 months ahead
  • Output: probability of recession one year into the future

Model training uses a walk-forward expanding window to mimic real-time forecasting.

---

Scenario-Based Probability Bands (Explained)

The probability bands shown in:

• scenario-based_conditional_recession_probabilities.png

represent conditional recession probabilities under different macro-economic paths, not Monte Carlo simulation.

Specifically:

1. Yield curve variables are held at their historical/observed values.

2. Macro inputs are set to different “scenarios”—

   • Soft landing path (strong labor market + improving credit spreads)
   • Baseline path
   • Downside path (weaker ISM, rising claims, widening credit spreads)

3. The logistic model is re-evaluated under each scenario to generate:

   • upper/lower probability bands
   • scenario-conditional trajectories
   • insights into when the yield curve signal strengthens or weakens

This allows the model to answer:

“What is the recession probability if macro conditions stay strong?
What if they deteriorate?
How wide is the uncertainty band?”

It is scenario conditioning, not Monte Carlo.

---

What Is Not Included Here

To protect intellectual property and avoid trivial duplication, the following are omitted:

• raw or processed data
• feature engineering code
• full model-training pipelines
• expanding-window validation logic
• configuration files
• model artifacts

The full implementation is stored in a private repo and available only for professional review.

---

Access & Contact

For collaboration, code review, or discussion of the full modeling framework:

Email: thebaselpoint@gmail.com
Substack: https://thebaselpoint.substack.com

© 2025 The Basel Point Research. All rights reserved.
Prepared by S.Y. Kim.