tests/test_risk_attribution.py

import sys
import os
import pytest
import numpy as np
import pandas as pd

_this_dir = os.path.dirname(os.path.abspath(__file__))
sys.path.insert(0, _this_dir)

from risk_attribution import cvar_attribution, stress_correlation

def test_cvar_fallback_branch():
    rng = np.random.default_rng(42)
    # Generate an artificially small dataset with very low variance
    # to trigger the "Empirical tail too thin (<3 obs)" fallback branch
    tickers = ["AAPL", "TLT"]
    dates = pd.date_range("2023-01-01", periods=10, freq="B")
    
    # Only 10 observations, tail at 95% will have 0.5 obs < 3
    returns_df = pd.DataFrame({
        "AAPL": rng.normal(0, 0.01, 10),
        "TLT": rng.normal(0, 0.01, 10)
    }, index=dates)
    
    weights = pd.Series({"AAPL": 0.6, "TLT": 0.4})
    
    # This should trigger the parametric fallback
    component_cvar, total_cvar = cvar_attribution(weights, returns_df, alpha=0.95)
    
    assert total_cvar > 0
    assert len(component_cvar) == 2
    assert "AAPL" in component_cvar
    
def test_stress_correlation_bounds():
    rng = np.random.default_rng(42)
    tickers = ["A", "B", "C"]
    
    # Base covariance matrix with some positive correlation
    cov = np.array([
        [0.04, 0.01, 0.01],
        [0.01, 0.04, 0.01],
        [0.01, 0.01, 0.04]
    ])
    cov_df = pd.DataFrame(cov, index=tickers, columns=tickers)
    weights = pd.Series({"A": 0.4, "B": 0.4, "C": 0.2})
    
    # Apply a massive shock to trigger clipping
    stressed_cov_df, stressed_vol = stress_correlation(weights, cov_df, shock_corr=0.9)
    
    # Check that correlations don't exceed 1.0 (implied by variance and covariance)
    vols = np.sqrt(np.diag(stressed_cov_df.values))
    outer_vols = np.outer(vols, vols)
    corr_mat = stressed_cov_df.values / outer_vols
    
    # Max correlation off-diagonal should be <= 1.0
    np.testing.assert_array_less(corr_mat - 1e-5, 1.0)
    
    # Stressed vol should be higher than normal vol
    normal_vol = np.sqrt(weights.values.T @ cov_df.values @ weights.values)
    assert stressed_vol > normal_vol