Abstract
This article presents and implements the main numerical methods used to replicate the Bovespa Index. The implemented models were full replication, minimum global variance portfolio, Black model and square minimization with short selling. The results were verified under the parameters of tracking error, beta, R² and serial similarity of average and variance. It was concluded that there is not the only way to achieve the best result, because all methods present strengths and weakness on the different contexts tested. In addition, managers may have different necessities on the time.
Replication models; tracking error; index arbitrage; square optimization; proxy portfolio