Index tracking and enhanced indexing using cointegration and correlation with endogenous portfolio selection

This article investigates the out-of-sample performance of cointegration and correlation methods for index tracking (IT) and enhanced indexing (EIT) strategies applied to Brazilian and U.S. market data. Our goal is to compare both methods as we strongly explore cointegration in relation to previous studies: we make the portfolio selection endogenous to the problem for this approach. The tests are performed using data from 2004 to 2014 with samples of 57 assets for Brazilian data and 96 assets for U.S. data; portfolios are built using combinations of at most 10 of these assets. Despite the extensive tests carried out, the overall result shows similar performance for both methods. For IT in the Brazilian market, there was a trade-off between better tracking error and higher turnover for cointegration (with the opposite for correlation), this pattern was not clear in the U.S. market. The outcome for the EIT also does not clearly favor cointegration or correlation.