Sector Correlation Matrix

How closely the 11 SPDR sector ETFs move together, trailing ~2 years of daily returns. The full pairwise matrix, how the average correlation drifts over time, and its relationship to market volatility.
Loading...
Current Avg. Pairwise Correlation
…
Trailing 63 trading days
Most Correlated Pair
…
Full ~2-year period
Most Diversifying Pair
…
Lowest pairwise correlation
Correlation vs. Volatility
…
r, avg. correlation vs. SPY realized vol

Full-period pairwise correlation matrix

Pearson correlation of daily returns between each pair of sector ETFs, trailing ~2 years

Average pairwise correlation over time

Rolling 63-trading-day average across all 55 sector pairs. Rises when sectors move together (broad risk-on/risk-off), falls when sector-specific stories dominate

Average correlation vs. SPY realized volatility

Each point is one trading day: x-axis is that day's trailing-63-day annualized SPY realized volatility, y-axis is that day's trailing-63-day average pairwise sector correlation. Testing the "correlations go to 1 in a crisis" claim directly

Correlation to the market

Each sector's correlation to SPY, full ~2-year period. Lower means more diversifying against a broad market position
SectorCorr. to SPY
Loading…

Methodology

Reuses the same daily adjusted-close history already fetched for Sector Performance (the 11 SPDR sector ETFs plus SPY, ~2 years of daily bars) rather than pulling any new Alpha Vantage data. This page is pure client-side analysis of an existing dataset. Daily % returns are computed from those closes, and every pairwise Pearson correlation is computed directly from the aligned daily return series (no smoothing or resampling).

The rolling correlation series takes a 63-trading-day (~one calendar quarter) trailing window, recomputes the average of all 55 sector-pair correlations at each day, and slides that window forward one day at a time. The scatter panel pairs that same rolling series against a rolling 63-day annualized realized volatility of SPY (standard deviation of daily returns × √252), then runs the same Pearson/Spearman regression check used throughout Factor Analysis. Note that adjacent days in a rolling-window series share almost all of their underlying data, so this regression's n overstates how many truly independent observations exist. Treat the r/p values as a rough gauge of direction and strength, not as a rigorous hypothesis test the way the non-overlapping-annual checks on other pages are.

The heatmap intentionally excludes SPY itself (an average of the sectors, not a peer of them). The market-relationship table below the heatmap covers each sector's own correlation to SPY separately.

Source: Yahoo Finance (via the existing /api/sector-performance dataset), TIME_SERIES_DAILY_ADJUSTED