The Intraday Correlation Pattern 

Authors:

Abstract
We establish a functional central limit theorem for a nonparametric estimator of the intraday periodic component of the instantaneous correlation between two financial assets, constructed from high-frequency data over a long time span. Our theory is developed in the Skorokhod space, which we show to be necessary, as the correlation functional fails to be Hadamard differentiable in the conventional L 2 metric; our approach further accommodates correlation patterns of either sign within the day as well as the benchmark case of a zero integrated correlation. We propose an estimator of the limiting covariance function, which enables a test for the equivalence of intraday correlation periodicity across two non-overlapping time periods. An empirical application to S&P 500 index and US Treasury bond futures uncovers pronounced intraday periodicity in correlation, whose shape varies over time, offering auxiliary information about potential drivers behind the long-term variation in the stock-bond correlation.