Arbitragem estatística em estratégias de long-short: as metodologias de cointegração e de correlação condicional dinâmica aplicadas ao mercado acionário brasileiro

Abstract

This dissertation investigates the performance of cointegration and dynamic conditional correlation methodologies applied in the elaboration of long-short strategy in the Brazilian stock market. The cointegration methodology was approached through the Engle and Granger test. Furthermore, the DCC-GARCH model was used to apply the dynamic conditional correlation methodology. Assessing the ability to generate surplus income through arbitrage statistics for each approach, a comparative performance analysis will be performed. For this purpose, the closing prices of shares listed on the IBrX 100 (Brazil 100 Index) were collected. Tradings related to formed pairs cover the period from 2008 to 2020. The results of this research demonstrate the ability to generate positive returns for cointegrated pairs and for pairs formed by Dynamic Conditional Correlation methodology. Comparatively, the pairs formed by the DCC-GARCH model performed higher cumulative returns than the cointegrated pairs, as well as showing greater capacity to surpass the appreciation of the IBOVESPA (Sao Paulo Stock Exchange Intex) for the period considered.