Time varying correlation: A key indicator in finance

Abstract

Correlations between different asset returns represent a crucial element in assets allocation decisions and financial engineering. In commodity markets, where prices result non stationary and returns are only mean stationary, a time varying measure of correlation has to be used. According to the prevailing literature, correlations among different markets are higher during recessions than during expansion periods. Portfolio managers to shield investors from stock markets declines used to invest in commodities which historically were considered poorly correlated with stock markets and providing a good hedge in the long run. In the last decade correlations between commodities and stock returns have dramatically changed. The aim of the paper is to address the issue of the correlation measurement in presence of non stationarity and structural breaks in market variables. We compare the Historical Rolling Correlation and the Dynamic Conditional Correlation methods and show how each estimator may provide useful information given a specific structure of the data. Some interesting relationships are also highlighted among markets where no correlations were expected and viceversa. We also show that information provided by the correlation measures can be used to identify structural breaks in the original variables.