Publicação
Modelling causality in nonstationary variances with an application to carbon markets
| Resumo: | In this paper we propose a multivariate generalisation of the multiplicative decomposition of the volatility within the class of conditional correlation GARCH models. The GARCH variance equations are multiplicatively decomposed into a deterministic nonstationary component describing the long-run movements in volatility and a short-run dynamic component allowing for volatility spillover effects across markets or assets. The conditional correlations are assumed to be time-invariant in its simplest form or generalised into a flexible dynamic parameterisation. Parameters of the model are estimated equation-by-equation by maximum likelihood applying the maximisation by parts algorithm to the variance equations, and thereafter to the structure of conditional correlations. An empirical application using carbon markets data illustrates the usefulness of the model. Our results suggest that, after modelling the variance equations accordingly, we find evidence that the transmission mechanism of shocks persists which is supported by the presence of variance interactions robust to nonstationarity. |
|---|---|
| Autores principais: | Martins, Susana |
| Outros Autores: | Amado, Cristina |
| Assunto: | Variance interactions Nonstationarity Short- and long-term volatility Lagrange multiplier test |
| Ano: | 2023 |
| País: | Portugal |
| Tipo de documento: | working paper |
| Tipo de acesso: | acesso aberto |
| Instituição associada: | Universidade do Minho |
| Idioma: | inglês |
| Origem: | RepositóriUM - Universidade do Minho |
| Resumo: | In this paper we propose a multivariate generalisation of the multiplicative decomposition of the volatility within the class of conditional correlation GARCH models. The GARCH variance equations are multiplicatively decomposed into a deterministic nonstationary component describing the long-run movements in volatility and a short-run dynamic component allowing for volatility spillover effects across markets or assets. The conditional correlations are assumed to be time-invariant in its simplest form or generalised into a flexible dynamic parameterisation. Parameters of the model are estimated equation-by-equation by maximum likelihood applying the maximisation by parts algorithm to the variance equations, and thereafter to the structure of conditional correlations. An empirical application using carbon markets data illustrates the usefulness of the model. Our results suggest that, after modelling the variance equations accordingly, we find evidence that the transmission mechanism of shocks persists which is supported by the presence of variance interactions robust to nonstationarity. |
|---|