Spliced Correlation : Theory Development

Correlation is a common metric used in portfolio management. It describes the relative movement of two streams of data, allowing inference of how one will behave given the movement of the other. Often a significant correlation relationship (whether it be uncorrelated, positively correlated or negatively correlated) in the long-term is not replicated in shorter term periods. Worse, often the short-term correlation is contradictory to the long-term. Utilizing three sets of data, where the streams of two are interchanged to form one stream at varying points of time could allow the long-term correlation to be also replicated in the short-term. There remain various obstacles to overcome, such as scaling, determination of inflection points and the selection of the data streams. Those are left to be solved in future papers - this paper puts forth the theoretical justification for the concept of spliced correlation