Financial Stress Composite

Abstract

The Financial Stress Composite blends three independent indicators of U.S. financial strain: the St. Louis Fed Financial Stress Index (STLFSI4), high-yield option-adjusted spreads (HY OAS), and the yield-curve slope (T10Y2Y). Higher stress corresponds to elevated STLFSI4, wider credit spreads, and deeper yield-curve inversion. Inputs are transformed to a common direction, robustly normalised using rolling median/MAD z-scores, and aggregated into a continuous composite that maps to High_Stress, Neutral, or Low_Stress.

1. Data

All inputs are sourced from FRED and resampled to month-end for consistent cross-signal comparison. Where a series is higher-frequency (daily/weekly), month-end uses the last available observation in the month.

2. Transformation

Stress rises with higher STLFSI4 and wider HY OAS, but tends to rise as the curve becomes more inverted. To align directions, invert the slope:

This makes higher values indicate more inversion and therefore higher “stress contribution.”

3. Normalisation

Each component is transformed into a robust rolling z-score to reduce outlier sensitivity.

4. Composite construction

The v1.0 composite used equal weights. In v1.1, the default remains equal-weight, but the framework supports regime-aware dynamic weights (Section 7–8).

Default: w_{STL,t}=w_{HY,t}=w_{INV,t}=1/3.

5. Regime mapping

Thresholds are intentionally coarse to emphasise regime classification over point forecasts.

6. Data freshness, staleness, and quality flags (v1.1)

Some underlying market series can be stale around holidays, data outages, or vendor delays. Because STLFSI4 is itself a composite that may embed volatility (VIX-like) and funding-spread (TED-like) inputs, staleness can propagate into the composite.

These flags do not change the historical values; they contextualise reliability of the latest print.

7. Sub-component contributions (v1.1)

To reduce interpretability gaps, publish each component’s contribution to the composite:

This makes it explicit whether “stress” is being driven by credit, the yield curve, or broad multi-market strain.

8. Dynamic weighting by market regime (v1.1)

v1.1 introduces an optional regime-aware weighting overlay intended to reduce double-counting and improve robustness when one pillar dominates. The objective is not optimisation, but sensible re-weighting when stress is clearly credit- driven vs curve-driven.

If you prefer a data-driven scheme, an alternative is to set w_{i,t} inversely proportional to each component’s rolling correlation with STLFSI4 (to downweight overlap) while keeping weights bounded (e.g., 0.2–0.5).

9. Implementation (Python)

import pandas as pd
import numpy as np

def robust_z(s, win=60, min_win=24):
    x = pd.to_numeric(s, errors="coerce").astype(float)
    med = x.rolling(win, min_periods=min_win).median()
    mad = (x - med).abs().rolling(win, min_periods=min_win).median()
    return (x - med) / (1.4826 * mad.replace(0, np.nan))

# Expect inputs: STLFSI4 (weekly), BAMLH0A0HYM2 (daily), T10Y2Y (daily)
# All resampled to month-end before applying robust_z.
f = df_fin_stress.copy()

# Transformations
f["Z_STL"] = robust_z(f["STLFSI4"])
f["Z_HY"]  = robust_z(f["HY_OAS"])
f["Z_INV"] = robust_z(-f["T10Y2Y"])   # inversion-aligned

# Default equal weights
w_eq = {"STL": 1/3, "HY": 1/3, "INV": 1/3}

# Optional regime-aware weights (simple example using lagged composite)
hi, lo = 0.75, -0.75
w_hi = {"STL": 0.40, "HY": 0.40, "INV": 0.20}

# First pass composite (equal-weight) to define the lagged regime
f["C_eq"] = (w_eq["STL"]*f["Z_STL"] + w_eq["HY"]*f["Z_HY"] + w_eq["INV"]*f["Z_INV"])

def weights_for_row(c_lag):
    if pd.isna(c_lag):
        return w_eq
    if c_lag > hi:
        return w_hi
    return w_eq

w_stl, w_hy, w_inv = [], [], []
for c_lag in f["C_eq"].shift(1):
    w = weights_for_row(c_lag)
    w_stl.append(w["STL"]); w_hy.append(w["HY"]); w_inv.append(w["INV"])

f["W_STL"], f["W_HY"], f["W_INV"] = w_stl, w_hy, w_inv

# Final composite and contributions
f["Fin_Stress"] = f["W_STL"]*f["Z_STL"] + f["W_HY"]*f["Z_HY"] + f["W_INV"]*f["Z_INV"]
f["Contrib_STL"] = f["W_STL"]*f["Z_STL"]
f["Contrib_HY"]  = f["W_HY"]*f["Z_HY"]
f["Contrib_INV"] = f["W_INV"]*f["Z_INV"]

def regime(v):
    if pd.isna(v):
        return np.nan
    return "High_Stress" if v > hi else ("Low_Stress" if v < lo else "Neutral")

f["Fin_Stress_Regime"] = f["Fin_Stress"].apply(regime)

df_sig_fin_stress = f

10. Interpretation (how to read the signal)

Academic and policy literature supports the core components: the term spread is a widely used leading indicator for recession risk and policy stance, credit spreads contain information about risk premia and predict real activity, and financial-stress indices aggregate cross-market strains into a single stress gauge.

11. Limitations

• Overlap risk: STLFSI4 already aggregates multiple spreads/volatility inputs; adding HY OAS and inversion can double-count certain dynamics.
• Monthly resampling: month-end snapshots can miss intramonth stress spikes; the signal is designed for regime context, not event detection.
• Threshold stability: fixed thresholds (±0.75) may under/over-state stress in structurally different volatility regimes; adaptive thresholds can be considered for trading use.
• Staleness: quality flags improve interpretability, but do not “fix” stale data; operational monitoring is still required.

12. References (interpretation anchors)

• FRED: St. Louis Fed Financial Stress Index (STLFSI4). Series page.
• FRED Blog: The St. Louis Fed’s Financial Stress Index, version 4. Article.
• FRED: ICE BofA U.S. High Yield Index Option-Adjusted Spread (BAMLH0A0HYM2). Series page.
• Estrella, A. & Mishkin, F. (1998): “Predicting U.S. recessions: Financial variables as leading indicators.” (NY Fed discussion summarized in yield-curve resources).
• New York Fed: Yield Curve as a Leading Indicator / recession probability model. Resource.
• Gilchrist, S. & Zakrajšek, E. (2012): “Credit Spreads and Business Cycle Fluctuations.” AER.