Transmissão de volatilidade da soja em Mato Grosso : uma aplicação de modelos GARCH

Abstract

This study analyzes the price volatility of soybeans across different local markets in the state of Mato Grosso, Brazil, and its interdependence with the national reference index provided by CEPEA. To this end, econometric models from the GARCH family were employed, including univariate ARMA-GARCH and multivariate approaches such as DCC-GARCH and BEKK-GARCH. Daily price series from May/2022 to April/2025 were transformed into log-returns to ensure comparability and stationarity. The univariate analysis revealed conditional heteroskedasticity, justifying the use of GARCH models. The multivariate models enabled the estimation of dynamic conditional correlations, the Total Correlation Spillover (TCS) index, and the net volatility transmission across markets. Results show that the CEPEA index operates as a price-setting market and a net transmitter of volatility to local markets. Furthermore, municipalities such as Diamantino and Sorriso played an active role in propagating shocks. The findings indicate strong integration between regional and national markets, with important implications for hedging strategies, price formation, and public policy aimed at stabilizing the agricultural sector.