Fractional Integration in Commodity Futures Returns

Abstract:     
We reexamine commodity futures returns for evidence of fractional integration utilizing two estimators based on wavelets. We summarize basic wavelet methods for signal processing and decompose commodity futures returns by wavelet scale. We find the evidence for long memory is not conclusive based on visual inspection of the wavelet decomposition, but formal statistical tests suggest evidence of long memory, in the form of anti-persistence, in about half of agricultural commodity futures. We find little evidence of long memory in metal futures. Our results are useful in interpreting previous disparate findings based on frequency domain estimators.

Keywords: futures returns, fractional integration, long memory, wavelets

JEL Classifications: G10, Q14

Accepted Paper Series