Essays on multivariate GARCH models and financial interactions

In this thesis, new modelling frameworks for analysing the dynamics of comovements of returns and volatility of returns in multivariate financial systems are proposed. The models considered belong to the class of multivariate conditional correlation GARCH models. The conditional volatilities are modelled as univariate GARCH processes to capture the heteroskedastic behaviour of the time series and the correlations are (conditionally) obtained for the (estimated) volatility standardized returns. Given the non-stationary nature of the variances for long time series and the occurrence of structural changes during crisis periods, conditional volatilities are described as non-stationary GARCH processes. The long-run volatility is modelled as a deterministic function where time is used as the transition variable. The structure of the time-varying correlations is chosen according to the purpose of the analysis and supported by misspecification tests. The first essay provides the first empirical application of a multivariate GARCH model with multiplicative decomposition to modelling comovements of changes in government bond yields. The presence and extent of sovereign debt contagion are analysed for nine countries of the euro area in the period between 2007 and 2017. An appealing feature of the new approach to contagion based on smooth transitions in the conditional correlations is the ability to distinguish long-run and short-run contagion effects based on the indicator for the time-variation in correlations. The findings indicate the presence of long-run and short-run contagion, respectively, during and after the more acute phase of the sovereign debt crisis across peripheral countries. The new additive time-varying structure of the GARCH equations introduced in the second essay is obtained by adding a time-dependent component to the vector GARCH process augmented with cross-market ARCH effects. In this setting, time-dependence in volatilities and co-volatilities is captured by smoothly changing deterministic functions of time. Contagion can be identified from crisis-contingent structural changes in the volatility interactions. Results show stronger comovements of the volatilities of the Greek, Irish and Portuguese bond markets during the recent sovereign debt crisis in the euro area providing empirical evidence for the presence of also volatility-based contagion. A new multivariate GARCH model with non-stationary GARCH equations is suggested in the last essay. The model with multiplicative decomposition of volatilities is able to capture both short- and long-term volatility dynamics. The modelling of the short-term volatility interactions is done after accounting for structural changes in the long-term volatilities. The conditional variance subcomponent describes heteroskedasticity and the volatility interactions in the short run. The subcomponent described by a deterministic function of time introduces non-stationarity and is used as a measure of volatility for the long run. An empirical application to major exchange rates against the euro indicates the presence of nonstationarities and stronger comovements in the conditional volatilities during the dot-com bubble and the recent global financial crisis.

Name thesis:degree_name

Tese de doutoramento em Economia

Grantor

Universidade do Minho

Year dc:date.issued

2019

Author dc:creator

• Martins, Susana Campos

Advisor dc:contributor.advisor

• Amado, Cristina

Statement dc:rights

• restrictedAccess

Language dc:language.iso

por

Handle dc:identifier.uri

https://hdl.handle.net/1822/65251

OAI record oai:identifier

http://repositorium.sdum.uminho.pt/oai/request?verb=GetRecord&metadataPrefix=dim&identifier=oai%3Arepositorium.uminho.pt%3A1822%2F65251