Análise da volatilidade de ativos financeiros na América Latina: uma perspectiva sobre as estruturas de dependência e otimização de carteiras

Abstract

Considering the risk-return relationship that investors are subject in their investment decisions, as well as the regulators in the development of the regulation for the stability of the financial markets, the present study analyzed the dependence of volatility between the economies of Latin America in periods of bear and bull markets, from copula models. In addition, it studied the empirical evidence that benefits may be gained by investing in these markets using time-varying returns and volatility forecasts from an ADCC-GARCH copula model. To do so, the official indices of five Latin American markets (Brazil, Mexico, Chile, Argentina and Peru), which it was measured in dollars to mitigate the effect of inflation and for comparability, were analyzed during the period from January 3, 2000 to 28 of December of 2017. The results showed that the most appropriate copula to modelling the structure of dependence of the markets was the symmetric Joe Clayton with variable parameters in the time. The dependence volatility structure was higher in the positive (upper-tail) than in the negative (lower-tail) returns, which may indicate that the Latin American markets during the analyzed period had diversification advantages during downturns due to less volatility of dependency among Latin American markets. The results also showed that the markets coupling in times of global crises, in the case of the subprime crisis than at other times; and also the existence of monetary and temporal effects in dependency structures. When analyzing the relationship of the indices in a multivariate structure for the optimization of a portfolio, it was shown that the Copula-ADCC-GARCH model demonstrates the smallest loss of information in terms of AIC and BIC. Further, we tested adherence of the data to the values performed in an analysis out-of-sample of returns and variance that it showed to be appropriate. The results showed that when looking for a strategy that seeks to optimize the relationship of lower variance (risk) and to achieve gain in terms of return on benchmarking, outperforming the naive strategy, it was confirmed that the investor can obtain gains in a multivariate structure with copulas distribution. This study demonstrated that when taking into account aspects of the financial market that were being neglected for the portfolio allocation, we can obtain gains in the risk-return relationship and be able to check existing patterns in the markets, which could bring benefits to investors and regulators.