How Highly Correlated Can Products Be Without Causing Problems?
(3/5 from 1 vote)
I am trying to optimize a portfolio for which a number of the products turn out to be highly correlated (greater than 99%). My approach has been to delete one of a highly correlated pair to leave whichever product shows the least correlation with the other products in the portfolio. How far do I need to carry this process? I started out with 14 products and removed 2 due to high correlation. Correlations in excess of 95% remain. My concern has to do with very small values amplifying round off error and possible near divide by zero if my products are highly correlated. This could cause the results to be unreliable. Do you have any rule of thumb regarding how correlated is too correlated?
There is no rule of thumb regarding correlation magnitude that we know of. Logically this would depend on the other correlations within the matrix. If two products are highly correlated but negatively correlated to other products, one would expect them to be weighted accordingly to their independent risk and return dynamics accordingly under the maximum Sharpe ratio for the portfolio. The correlation is only one attribute contributing to the optimal weighting as average return and standard deviation of returns for each product are also factors.