CONVEX
Correlation Deep Dive

Food CPI vs Energy CPI: Correlation Analysis

Pearson correlation of daily returns for CPI: Food and CPI: Energy. Rolling windows, yearly breakdown, regression beta, and divergence analysis. Data window spans to (56 aligned observations).

30-Day
-0.309
Weak negative
90-Day
+0.163
Essentially uncorrelated
1-Year
+0.163
Essentially uncorrelated
5-Year
+0.163
Essentially uncorrelated

What the Number Means

With a correlation of 0.16, CPI: Food and CPI: Energy are essentially uncorrelated at daily frequency. Either the relationship operates at a different time horizon or the shared driver has been dominated by idiosyncratic noise during the observation window.

Recent vs Long-Run Behavior

Last 90 Days
+0.163
5-Year Baseline
+0.163

Recent correlation tracks the long-run relationship closely. No meaningful divergence. The historical pattern between CPI: Food and CPI: Energy is intact and should continue to serve as a reasonable baseline for positioning.

Statistical Details (1-Year Window)

Pearson Correlation (r)+0.163
R-Squared (r²)0.027
Beta (CPI: Food vs CPI: Energy)0.017
Daily Volatility σ(CPI: Food)0.31%
Daily Volatility σ(CPI: Energy)2.91%
Observations56

Correlation measures directional co-movement; R² quantifies the fraction of variance explained by the linear relationship. Beta is the slope coefficient from regressing CPI: Food returns on CPI: Energy returns. A beta above 1 means the first asset amplifies moves of the second.

Year-by-Year Correlation

YearCorrelationStrengthObservations
2026Insufficient data3
2025+0.014Essentially uncorrelated11
2024-0.259Weak negative12
2023+0.183Essentially uncorrelated12
2022+0.324Weak positive12
2021+0.653Strong positive6

Year-by-year correlation reveals how the relationship has held up across different macro regimes. Sharp year-over-year swings in correlation often mark the transition between stress and calm periods.

Methodology

Correlations are computed on daily log-adjacent returns for CPI: Food and CPI: Energy, aligned on shared trading dates. We use the Pearson product-moment coefficient, which measures the linear relationship between two return series.

Windows are the most recent N observations for 30D, 90D, and 1Y (252 trading days); the 5Y figure uses all aligned data up to 1,260 observations. Beta is the OLS slope from regressing the first series on the second. Data updates daily with a 24-hour revalidation cadence.

Related Correlations

More Comparisons

Get daily macro analysis on shifting correlations, regime transitions, and cross-asset signals.

Correlation is not causation and backward-looking statistics can fail when regimes shift. Positions sized on historical correlation assumptions should be stress-tested against scenarios where the relationship breaks. For informational purposes only.