Rate Sensitivity & Regime Dashboard

Core Question

How does bond duration risk behave across different yield curve regimes?

Why This Project Exists

Duration risk is not static. The same 10-year bond carries different practical risk depending on whether the yield curve is inverted, re-steepening, flat, or normal. This dashboard quantifies that relationship empirically — using historical FRED data to show how yield volatility, directional bias, and approximate price sensitivity have varied across regime states.

Data Sources

All data from the Federal Reserve Economic Data (FRED):

Series	Description	Frequency
DGS2	2-Year Treasury Constant Maturity	Daily → monthly
DGS10	10-Year Treasury Constant Maturity	Daily → monthly
DGS30	30-Year Treasury Constant Maturity	Daily → monthly
FEDFUNDS	Effective Federal Funds Rate	Monthly

All series resampled to month-end frequency. Short gaps (≤ 3 months) are forward-filled. DGS30 was discontinued Feb 2002 – Feb 2006; that gap is preserved as NaN and excluded from 30Y statistics.

Methodology

Regime Classification

See Regime Definitions below. Same thresholds as companion project yield-curve-inflation-dashboard.

Yield Change Statistics

Monthly first differences of each yield series (ppts). Average and standard deviation computed per regime, converted to basis points for display.

DV01 Proxy — Important Disclaimer

This is a simplified educational approximation. It is not suitable for risk management, trading, or investment decisions.

Formula:

modified_duration ≈ maturity × 0.85   (DURATION_FACTOR)
dv01_proxy        = −modified_duration × (monthly_yield_change / 100)

Result: approximate fractional price change per month (×100 for %). The 0.85 factor is a rough proxy for the relationship between maturity and modified duration across a range of coupon levels. Real-world DV01 requires:

• Actual coupon and settlement terms
• Day-count conventions
• Full yield-to-price mapping (especially non-trivial for large moves)
• Convexity adjustment for large yield changes

The 30Y proxy (modified duration ≈ 25.5) will show the largest price sensitivity, which is directionally correct but numerically approximate.

Directional Bias Analysis

For each regime: percentage of months where 10Y yields rose vs fell, and average magnitude in each direction. The point is that regimes differ not just in volatility magnitude but in directional tendency.

Rolling 12-Month Adverse Moves

Rolling 12-month cumulative yield change computed for each maturity. The 90th percentile of adverse (upward) moves is shown per regime — representing a historically significant but not extreme rate stress within that regime.

Regime Definitions

Yield curve regimes are based on the 10Y–2Y spread, applied in strict priority order:

1. Re-steepening (highest priority) — ALL must hold:
   • Spread was negative at any point in prior 6 months
   • Spread has risen > 0.25 ppts over last 3 months
   • Current spread is between −0.25 and +0.75 ppts
2. Inverted — spread < 0
3. Flat — spread 0 to 0.50 ppts
4. Normal — spread > 0.50 ppts

Dashboard Walkthrough

Section	Content
1 — Current Snapshot	2Y, 10Y, 30Y yields; Fed Funds; spread; current regime; 2-sentence context
2 — Yield Volatility	Grouped bar: avg change & σ by regime for 2Y and 10Y
3 — Duration Sensitivity	DV01 heatmap + directional bias table + adverse moves chart
4 — Risk Summary	4 data-driven bullets: highest volatility, most exposed maturity, directional bias, current regime
5 — Regime → Portfolio Playbook	Heuristic interpretation layer: volatility label, directional bias, regime character, duration risk, carry environment, convexity value, and positioning context bullets — all derived rule-based from historical regime statistics

Key Takeaways

• Inverted regimes have historically exhibited the highest 2Y yield volatility as policy rate expectations shift.
• 30Y duration amplifies adverse moves significantly relative to 2Y in all regimes — the DV01 heatmap makes this explicit.
• Re-steepening is rare (~4% of months) but has historically coincided with transition dynamics that compress term premium.
• Directional bias varies by regime: not all yield changes in a given regime are adverse — the % of months rising vs falling matters as much as magnitude.
• The Playbook section translates current regime into descriptive risk labels (duration, carry, convexity) and positioning context derived from historical regime statistics — not forecasts.

How to Run

cd rate-sensitivity-regime-dashboard
pip install -r requirements.txt
streamlit run app.py

Optional: FRED API Key

cp .env.example .env
# Edit .env: FRED_API_KEY=your_key_here

Free API key: https://fred.stlouisfed.org/docs/api/api_key.html

Data is cached locally in data/processed/ as parquet files on first run (refreshed every 24 hours).

Disclaimer

This is an educational portfolio project. It is not investment advice. DV01 values are simplified proxies using a modified duration approximation. All regime logic is rule-based and for analytical illustration only. Historical patterns are not predictive of future outcomes.