Stylized Facts of Financial Returns › Correlation Breakdown in Crises

Michael BrenndoerferNovember 21, 20253 min read
Quantitative FinanceAll chapters
Part 5 of 15Stochastic Calculus and Pricing Theory
Chapter 25Stylized Facts of Financial Returns › Correlation Breakdown in Crises
Reading section 9 of 14 · Correlation Breakdown in Crises · 3 minSection 9 of 14 · 3 min

Book contents

Section 9 of 14 · 3 min read

Stylized Facts of Financial Returns › Correlation Breakdown in CrisesOne section · Sections in this chapter9 / 14 ↓
Reading tools Definitions & inline links

Choose your expertise level to adjust how many terms are explained. Beginners see more tooltips, experts see fewer to maintain reading flow. Hover over underlined terms for instant definitions.

Article links

Make inline references clickable

A final stylized fact concerns correlations between assets, which increase during market stress. The diversification benefits you count on in normal times partially evaporate precisely when you need them most.

Even after controlling for volatility, the underlying correlation structure shifts during crises. Assets that seemed uncorrelated become highly correlated when panic sets in.

In[32]:
Code
# Define crisis and calm periods
crisis_periods = [
    ("2008-09-01", "2009-03-31"),  # Financial Crisis
    ("2020-02-15", "2020-04-15"),  # COVID Crash
]

calm_periods = [
    ("2013-01-01", "2013-12-31"),
    ("2017-01-01", "2017-12-31"),
]


def get_corr_matrix(returns, start, end):
    mask = (returns.index >= start) & (returns.index <= end)
    return returns[mask].corr()


# Calculate correlations for different periods
crisis_corr_2008 = get_corr_matrix(multi_returns, "2008-09-01", "2009-03-31")
calm_corr_2017 = get_corr_matrix(multi_returns, "2017-01-01", "2017-12-31")
crisis_corr_2020 = get_corr_matrix(multi_returns, "2020-02-15", "2020-04-15")
Out[33]:
Console
Correlation Structure: Crisis vs. Calm Periods
============================================================

Financial Crisis (Sep 2008 - Mar 2009):
       ^GSPC    GLD    TLT    EFA
^GSPC  1.000  0.394 -0.013  0.945
GLD    0.394  1.000  0.196  0.448
TLT   -0.013  0.196  1.000  0.037
EFA    0.945  0.448  0.037  1.000

Calm Period (2017):
       ^GSPC    GLD    TLT    EFA
^GSPC  1.000  0.063 -0.351  0.862
GLD    0.063  1.000  0.153  0.101
TLT   -0.351  0.153  1.000 -0.310
EFA    0.862  0.101 -0.310  1.000

COVID Crash (Feb - Apr 2020):
       ^GSPC    GLD    TLT    EFA
^GSPC  1.000  0.050 -0.156  0.928
GLD    0.050  1.000  0.178  0.088
TLT   -0.156  0.178  1.000 -0.069
EFA    0.928  0.088 -0.069  1.000

During the financial crisis and COVID crash, correlations between S&P 500 and other assets shift compared to calm periods. The correlation between US and international equities (EFA) typically spikes during crises, reducing the diversification benefit of global equity allocation exactly when losses are largest.

Out[34]:
Visualization
Heatmap of a 4x4 asset correlation matrix for SPX, GLD, TLT, and EFA during the calm 2017 period, with moderate to low off-diagonal values.
Correlation matrix of asset returns during a calm market period in 2017. Correlations between the S&P 500 and other asset classes are moderate to low, allowing for effective portfolio diversification.
Heatmap of a 4x4 asset correlation matrix for SPX, GLD, TLT, and EFA during the 2020 COVID crash, with noticeably higher off-diagonal correlations than the calm period.
Correlation matrix during the 2020 COVID-19 crash. The strongest equity pair, SPX and EFA, rises from 0.86 to 0.93, while other pairs change by different amounts and directions.