Credit Spreads

Abstract

We build a monthly Credit Spreads Composite (CSC) from ICE BofA US High Yield OAS and US Corporate Master OAS. Level and momentum features are robust‑standardised and combined into a composite that maps to regimes—EASY, NORMAL, TIGHTENING, STRESSED. The CSC feeds into the broader Financial Stress Index (FSI) as a key sub‑pillar alongside Liquidity, VIX, and Funding indicators.

Version 1.1 — Methodology Amendments

1) Data (FRED Identifiers)

• BAMLH0A0HYM2 — ICE BofA US High Yield Index Option‑Adjusted Spread (pct points).
• BAMLC0A0CM — ICE BofA US Corporate Master Option‑Adjusted Spread (pct points).

Frequency: daily; we aggregate to month‑end or week‑end snapshots. Units are percentage points (bps/100). Titles and last‑obs dates are logged.

2) Data Quality & Validation

data_quality_flag records staleness, ffill count, and window adaptations.

3) Component Transforms

For spread series s_t (pct points):

Constructed components for HY and Corp Master: hy_level, hy_d3m_ann, hy_d12m, hy_pct_rank, and analogues for ig_*.

4) Standardisation (Rolling z-scores)

Compute rolling z‑scores using robust median/MAD; fallback to mean/std if MAD=0 or short history.

5) Composite & Weighting

We emphasise the High Yield level and percentile, as HY is more sensitive to stress, with supporting weight from changes and IG spreads. Missing inputs are renormalised out.

{
  "hy_level_z": 0.30, "hy_pct_rank": 0.20, "hy_d3m_ann_z": 0.15, "hy_d12m_z": 0.10,
  "ig_level_z": 0.15, "ig_pct_rank": 0.05, "ig_d3m_ann_z": 0.03, "ig_d12m_z": 0.02
}

Experimental dynamic weighting (optional): compute component volatilities over a rolling window (e.g., 36–60 months) and tilt weights toward more stable components (inverse‑volatility) or toward HY components when a broader volatility regime is elevated (e.g., VIX/STLFSI high). Publish both CSC_fixed and CSC_dyn to preserve comparability while enabling research.

6) Scaling & Regime Mapping

Regimes communicate stress levels using composite z‑scores and absolute spread bands:

We also publish a 0–100 score via rolling min–max over 10 years.

Persistence / momentum overlay (v1.1): require that either (a) CSC z remains above the threshold for 2 consecutive observations, or (b) the level condition (HY) and momentum condition (Δ3m annualised) agree, before upgrading to TIGHTENING/STRESSED. This reduces whipsaws from one‑day prints or microstructure noise.

Time-varying percentiles: percentile ranks are computed on a rolling 10‑year window (120 months) by default and are recalculated each period; avoid comparing percentile ranks across different window settings without noting the window.

7) Integration with Financial Stress Index (FSI)

8) Implementation Notes (Python)

# FRED series (daily)
hy = fred.get_series("BAMLH0A0HYM2")   # HY OAS (pct points)
ig = fred.get_series("BAMLC0A0CM")     # Corp Master OAS (pct points)

# Resample to month-end with bounded ffill (<=5b)
hy_m = hy.resample("M").last().ffill(limit=5)
ig_m = ig.resample("M").last().ffill(limit=5)

def ann_3m(s): return 4*(s - s.shift(3))
components = {
  "hy_level": hy_m, "hy_d3m_ann": ann_3m(hy_m), "hy_d12m": hy_m - hy_m.shift(12),
  "ig_level": ig_m, "ig_d3m_ann": ann_3m(ig_m), "ig_d12m": ig_m - ig_m.shift(12),
}

# Percentile ranks over 120 months; robust rolling z-scores; weights -> CSC; map regimes.

9) Reproducibility & Monitoring

• Persist FRED IDs, retrieval timestamps, resampling policy, and windows.
• Log winsorisation bounds, staleness flags, and window adaptations.
• Store panels: raw, transforms, z-scores, contributions, composite, regimes.

10) Interpretation & Applications

Credit spreads price default and liquidity premia. Rising HY spreads typically precede equity drawdowns and slower capex. Use CSC in concert with VIX and Liquidity to calibrate risk appetite and stress testing.

11) Governance & Change Control

• Quarterly review of components, thresholds, and weights vs backtests.
• Document methodology updates; maintain semantic versioning.