Application of Copula-based Correlation Coefficients and Dynamic Programming-based Approaches to determining Similarity between Time Series for Clustering and Index Tracking

Index tracking that is known as one of the most attractive passive management methods of investing seeks to form portfolios in such a way that replicate the performance of market index as closely as possible. In this study, the application of a binary optimization model in time series clustering to form an index tracker portfolio is investigated. For the clustering process, various time series similarity measures including Capula-based correlation coefficients and dynamic distance based approaches (DTW and EDR) have been used. Out-of-sample test on market ratios and tracking error of portfolios based on 50 More Active Company index of Tehran Stock Exchange in the period from beginning of 1394 till the end of spring 1397 indicates that portfolios have been successful in tracking index and average tracking error of portfolios did not differ significantly. Also, pairwise comparisons test on portfolio tracking error indicates that portfolios tracking error does not differ significantly.