Volatilidade setorial: análise de causalidade e transbordamentos cambiais e financeiros

Abstract

The determination of cost of capital and investment and leverage decisions evaluation require understanding several factors, including volatility. The Brazilian stock market can be organized by sectors, which have different features in relation to the volatility behavior in the face of exchange rate and Ibovespa index shocks. This study uses daily return data from seven sector indexes, Ibovespa index and Dollar to analyze volatility spillovers through conditional correlations. Univariate GARCH models demonstrate that the return series of the sectoral indexes are characterized by the presence of heavy tails, skewness and kurtosis and were used as an indication of the multivariate models orders. Diagonal Multivariate GARCH-BEKK models with skewness parameters and low orders represented satisfactorily the sectoral indices generating series. Conditional correlations were estimated for three different models, combining the volatility of each sector with different shocks: dollar volatility, Ibovespa volatility and dollar and Ibovespa volatility. These correlations were also studied in periods with and without economic recession, indicating changes between the volatilities of the indexes in different scenarios. Second-order Granger causality tests were used to check for the presence of volatility spillovers and the directions, in order to characterize lead-lag effects. The results confirmed the existence of foreign exchange and Ibovespa spillovers for the sectoral indexes, as well as between the indexes. UTIL and IEE indexes showed several similarities. While the IMOB and INDX do not have movements anticipated by the volatility of other series, the IMAT occurs lagged, as it receives volatility spillovers of all series. The IFNC series showed characteristics similar to the Ibovespa. The average conditional correlations of IMOB and IFNC showed magnitude variations due to exchange rate shocks in periods with and without recession. Several other relationships are monstrated in this study.